before_code stringlengths 30 149k | reviewer_comment stringlengths 16 20.6k | after_code stringlengths 30 163k | diff_context stringlengths 0 22.4k | file_path stringlengths 6 155 | comment_line int64 0 26 | language stringclasses 19
values | quality_score float64 0.07 1 | comment_type stringclasses 9
values | comment_length int64 16 20.6k | before_lines int64 2 4.61k | after_lines int64 2 4.4k | is_negative bool 2
classes | pr_title stringlengths 7 140 | pr_number int64 18 61.2k | repo_name stringclasses 29
values | repo_stars int64 321 56k | repo_language stringclasses 14
values | reviewer_username stringclasses 378
values | author_username stringclasses 739
values | user stringlengths 361 200k | assistant stringlengths 16 20.6k | system stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
i -= 1
break
end
end
pathpart = join(path_arr[i+1:something(findlast(x -> !isempty(x), path_arr), 0)], path_separator)
prefix_num = something(findlast(x -> !isempty(x), start_arr), 0) - i - 1
if prefix_num >= 0
prefix = pardir * path_separator
relpath_ = i... | codeunits are not utf8. but we can easily do something about that:
```suggestion
for cu in utf8units(s)
``` | end
return isempty(relpath_) ? curdir : relpath_
end
relpath(path::AbstractString, startpath::AbstractString) =
relpath(String(path)::String, String(startpath)::String)
for f in (:isdirpath, :splitdir, :splitdrive, :splitext, :normpath, :abspath, :isabspath)
@eval $f(path::AbstractString) = $f(String(... | @@ -610,14 +777,28 @@ end
relpath(path::AbstractString, startpath::AbstractString) =
relpath(String(path)::String, String(startpath)::String)
-for f in (:isdirpath, :splitdir, :splitdrive, :splitext, :normpath, :abspath)
+for f in (:isdirpath, :splitdir, :splitdrive, :splitext, :normpath, :abspath, :isabspath)
... | base/path.jl | 26 | Julia | 0.714 | suggestion | 114 | 51 | 51 | false | Remove `Regex` / libpcre dependency from basic `path.jl` operations | 60,677 | JuliaLang/julia | 35,573 | Julia | vtjnash | johroj | Review this pull request change.
Language: Julia
File: base/path.jl
PR Title: Remove `Regex` / libpcre dependency from basic `path.jl` operations
BEFORE CODE: i -= 1
break
end
end
pathpart = join(path_arr[i+1:something(findlast(x -> !isempty(x), path_arr), 0)], path_separator... | codeunits are not utf8. but we can easily do something about that:
```suggestion
for cu in utf8units(s)
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
if path_arr[i] != start_arr[i]
i -= 1
break
end
end
pathpart = join(path_arr[i+1:something(findlast(x -> !isempty(x), path_arr), 0)], path_separator)
prefix_num = something(findlast(x -> !isempty(x), start_arr), 0) - i - 1
if prefix_num >= 0
prefix = pardi... | ```suggestion
out = empty!(StringVector(sizeof(s)))
``` | relpath_ = pathpart
end
return isempty(relpath_) ? curdir : relpath_
end
relpath(path::AbstractString, startpath::AbstractString) =
relpath(String(path)::String, String(startpath)::String)
for f in (:isdirpath, :splitdir, :splitdrive, :splitext, :normpath, :abspath, :isabspath)
@eval $f(path::... | @@ -610,14 +777,28 @@ end
relpath(path::AbstractString, startpath::AbstractString) =
relpath(String(path)::String, String(startpath)::String)
-for f in (:isdirpath, :splitdir, :splitdrive, :splitext, :normpath, :abspath)
+for f in (:isdirpath, :splitdir, :splitdrive, :splitext, :normpath, :abspath, :isabspath)
... | base/path.jl | 26 | Julia | 0.571 | suggestion | 61 | 51 | 51 | false | Remove `Regex` / libpcre dependency from basic `path.jl` operations | 60,677 | JuliaLang/julia | 35,573 | Julia | vtjnash | johroj | Review this pull request change.
Language: Julia
File: base/path.jl
PR Title: Remove `Regex` / libpcre dependency from basic `path.jl` operations
BEFORE CODE: if path_arr[i] != start_arr[i]
i -= 1
break
end
end
pathpart = join(path_arr[i+1:something(findlast(x -> !ise... | ```suggestion
out = empty!(StringVector(sizeof(s)))
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
(on Windows) has a drive specification that doesn't match the drive computed for
the join of the preceding paths, then prior components are dropped.
Note on Windows since there is a current directory for each drive, `joinpath("c:", "foo")`
represents a path relative to the current directory on drive "c:" so this is eq... | While this is slow path operations, I don't think we can suddenly decide to make it this inefficient | ```
```jldoctest
julia> joinpath(["/home/myuser", "example.jl"])
"/home/myuser/example.jl"
```
"""
joinpath
function _split_at_separators(path::AbstractString; keepempty = true)
# Equivalent to Base.split(path, r"/+"sa; keepempty) (r"[\\/]+"sa on windows)
# Since there is no split between consecutive separato... | @@ -372,6 +514,31 @@ julia> joinpath(["/home/myuser", "example.jl"])
"""
joinpath
+function _split_at_separators(path::AbstractString; keepempty = true)
+ # Essentially Base.split(path, isseparator; keepempty)
+ # with the extra condition that consecutive separators are
+ # treated as one separator.
+ chars::Vector... | base/path.jl | 26 | Julia | 0.357 | performance | 100 | 51 | 51 | false | Remove `Regex` / libpcre dependency from basic `path.jl` operations | 60,677 | JuliaLang/julia | 35,573 | Julia | vtjnash | johroj | Review this pull request change.
Language: Julia
File: base/path.jl
PR Title: Remove `Regex` / libpcre dependency from basic `path.jl` operations
BEFORE CODE: (on Windows) has a drive specification that doesn't match the drive computed for
the join of the preceding paths, then prior components are dropped.
Note on... | While this is slow path operations, I don't think we can suddenly decide to make it this inefficient | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
Join path components into a full path. If some argument is an absolute path or
(on Windows) has a drive specification that doesn't match the drive computed for
the join of the preceding paths, then prior components are dropped.
Note on Windows since there is a current directory for each drive, `joinpath("c:", "foo")`
... | This appears to be the definition of keepempty=false, so I'm not sure you can say it is "essentially the same as a flag" that it re-implements without that flag | "/home/myuser/example.jl"
```
```jldoctest
julia> joinpath(["/home/myuser", "example.jl"])
"/home/myuser/example.jl"
```
"""
joinpath
function _split_at_separators(path::AbstractString; keepempty = true)
# Equivalent to Base.split(path, r"/+"sa; keepempty) (r"[\\/]+"sa on windows)
# Since there is no split be... | @@ -372,6 +514,31 @@ julia> joinpath(["/home/myuser", "example.jl"])
"""
joinpath
+function _split_at_separators(path::AbstractString; keepempty = true)
+ # Essentially Base.split(path, isseparator; keepempty)
+ # with the extra condition that consecutive separators are
+ # treated as one separator. | base/path.jl | 26 | Julia | 0.357 | suggestion | 160 | 51 | 51 | false | Remove `Regex` / libpcre dependency from basic `path.jl` operations | 60,677 | JuliaLang/julia | 35,573 | Julia | vtjnash | johroj | Review this pull request change.
Language: Julia
File: base/path.jl
PR Title: Remove `Regex` / libpcre dependency from basic `path.jl` operations
BEFORE CODE: Join path components into a full path. If some argument is an absolute path or
(on Windows) has a drive specification that doesn't match the drive computed f... | This appears to be the definition of keepempty=false, so I'm not sure you can say it is "essentially the same as a flag" that it re-implements without that flag | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
julia> isabspath("home")
false
```
"""
isabspath(path::AbstractString)
"""
isdirpath(path::AbstractString)::Bool
Determine whether a path refers to a directory (for example, ends with a path separator).
# Examples
```jldoctest
julia> isdirpath("/home")
false
julia> isdirpath("/home/")
true
```
"""
function isd... | These newlines are a bit surprising.
They don't directly appear in the regex above, can you explain where they come from? | # Examples
```jldoctest
julia> isabspath("/home")
true
julia> isabspath("home")
false
```
"""
isabspath(path::AbstractString)
"""
isdirpath(path::AbstractString)::Bool
Determine whether a path refers to a directory (for example, ends with a path separator).
# Examples
```jldoctest
julia> isdirpath("/home")
fals... | @@ -133,7 +237,12 @@ julia> isdirpath("/home/")
true
```
"""
-isdirpath(path::String) = occursin(path_directory_re, splitdrive(path)[2])
+function isdirpath(path::String)::Bool
+ # Reimplements occursin(r"(?:^|/)\.{0,2}$"sa, splitdrive(path)[2])
+
+ _, after_last_separator = _splitdir_nodrive("", splitdrive(pa... | base/path.jl | 26 | Julia | 0.357 | question | 122 | 51 | 51 | false | Remove `Regex` / libpcre dependency from basic `path.jl` operations | 60,677 | JuliaLang/julia | 35,573 | Julia | topolarity | johroj | Review this pull request change.
Language: Julia
File: base/path.jl
PR Title: Remove `Regex` / libpcre dependency from basic `path.jl` operations
BEFORE CODE:
julia> isabspath("home")
false
```
"""
isabspath(path::AbstractString)
"""
isdirpath(path::AbstractString)::Bool
Determine whether a path refers to a ... | These newlines are a bit surprising.
They don't directly appear in the regex above, can you explain where they come from? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@inline isdriveletter(c::Char) = isascii(c) && isdriveletter(UInt8(c))
@inline function isdriveletter(c::UInt8)
UInt8('A') <= c <= UInt8('Z') ||
UInt8('a') <= c <= UInt8('z')
end
function _split_longunc(s::String)::Tuple{String, String}
# Long UNC path, e.g. `\\?\UNC\server\shar... | Can you comment to note where this `10` comes from? | @inline isseparator(c::UInt8) = c === UInt8('/') || c === UInt8('\\')
@inline isdriveletter(c::Char) = isascii(c) && isdriveletter(UInt8(c))
@inline function isdriveletter(c::UInt8)
UInt8('A') <= c <= UInt8('Z') ||
UInt8('a') <= c <= UInt8('z')
end
function _split_longunc(s::String... | @@ -20,40 +22,124 @@ export
if Sys.isunix()
const path_separator = "/"
- const path_separator_re = r"/+"sa
- const path_directory_re = r"(?:^|/)\.{0,2}$"sa
- const path_dir_splitter = r"^(.*?)(/+)([^/]*)$"sa
- const path_ext_splitter = r"^((?:.*/)?(?:\.|[^/\.])[^/]*?)(\.[^/\.]*|)$"sa
+ @inline... | base/path.jl | 26 | Julia | 0.357 | question | 51 | 51 | 51 | false | Remove `Regex` / libpcre dependency from basic `path.jl` operations | 60,677 | JuliaLang/julia | 35,573 | Julia | topolarity | johroj | Review this pull request change.
Language: Julia
File: base/path.jl
PR Title: Remove `Regex` / libpcre dependency from basic `path.jl` operations
BEFORE CODE: @inline isdriveletter(c::Char) = isascii(c) && isdriveletter(UInt8(c))
@inline function isdriveletter(c::UInt8)
UInt8('A') <= c <= UInt8('Z')... | Can you comment to note where this `10` comes from? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
start_arr = _split_at_separators(abspath(path_drive * startpath_without_drive))
else
path_arr = _split_at_separators(abspath(path))
start_arr = _split_at_separators(abspath(startpath))
end
i = 0
while i < min(length(path_arr), length(start_arr))
i += 1
if path_ar... | This is much better than `percent_escape` implicitly assuming it's called on a UTF-8 string, thanks | i -= 1
break
end
end
pathpart = join(path_arr[i+1:something(findlast(x -> !isempty(x), path_arr), 0)], path_separator)
prefix_num = something(findlast(x -> !isempty(x), start_arr), 0) - i - 1
if prefix_num >= 0
prefix = pardir * path_separator
relpath_ = i... | @@ -610,14 +773,28 @@ end
relpath(path::AbstractString, startpath::AbstractString) =
relpath(String(path)::String, String(startpath)::String)
-for f in (:isdirpath, :splitdir, :splitdrive, :splitext, :normpath, :abspath)
+for f in (:isdirpath, :splitdir, :splitdrive, :splitext, :normpath, :abspath, :isabspath)
... | base/path.jl | 26 | Julia | 0.429 | security | 99 | 51 | 51 | false | Remove `Regex` / libpcre dependency from basic `path.jl` operations | 60,677 | JuliaLang/julia | 35,573 | Julia | topolarity | johroj | Review this pull request change.
Language: Julia
File: base/path.jl
PR Title: Remove `Regex` / libpcre dependency from basic `path.jl` operations
BEFORE CODE: start_arr = _split_at_separators(abspath(path_drive * startpath_without_drive))
else
path_arr = _split_at_separators(abspath(path))
... | This is much better than `percent_escape` implicitly assuming it's called on a UTF-8 string, thanks | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@test splitdir(S("C:\\")) == ("C:\\", "")
@test splitdir(S("C:\\foo")) == ("C:\\", "foo")
@test splitdir(S("C:\\foo\\bar")) == ("C:\\foo", "bar")
@test splitdir(S("\\\\?\\C:\\foo\\bar")) == ("\\\\?\\C:\\foo", "bar")
@test splitdir(S("\\\\?\\C:\\foo\\bar\\")) =... | Now that we have our parity check complete (thanks for achieving that first), I think we can also fix-up this newline handling as well similar to @JeffBezanson 's suggestion above
| @test splitdir(S("C:\\")) == ("C:\\", "")
@test splitdir(S("C:\\foo")) == ("C:\\", "foo")
@test splitdir(S("C:\\foo\\bar")) == ("C:\\foo", "bar")
@test splitdir(S("\\\\?\\C:\\foo\\bar")) == ("\\\\?\\C:\\foo", "bar")
@test splitdir(S("\\\\?\\C:\\foo\\bar\\")) =... | @@ -200,6 +255,20 @@
@test_broken splitext(S(".foo..")) == (".foo", "..")
@test_broken splitext(S(".foo...")) == (".foo", "...")
@test splitext(S(".foo.bar")) == (".foo", ".bar")
+ @test splitext(S("bar/.foo/baz")) == ("bar/.foo/baz", "")
+ @test splitext(S("bar/foo/.baz")) == (... | test/path.jl | 26 | Julia | 0.357 | suggestion | 181 | 51 | 51 | false | Remove `Regex` / libpcre dependency from basic `path.jl` operations | 60,677 | JuliaLang/julia | 35,573 | Julia | topolarity | johroj | Review this pull request change.
Language: Julia
File: test/path.jl
PR Title: Remove `Regex` / libpcre dependency from basic `path.jl` operations
BEFORE CODE: @test splitdir(S("C:\\")) == ("C:\\", "")
@test splitdir(S("C:\\foo")) == ("C:\\", "foo")
@test splitdir(S("C:\\foo\\bar"... | Now that we have our parity check complete (thanks for achieving that first), I think we can also fix-up this newline handling as well similar to @JeffBezanson 's suggestion above
| You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
isdirpath,
joinpath,
normpath,
realpath,
relpath,
splitdir,
splitdrive,
splitext,
splitpath
if Sys.isunix()
const path_separator = "/"
const path_separator_re = r"/+"sa
const path_directory_re = r"(?:^|/)\.{0,2}$"sa
const path_dir_splitter = r"^(.*?)(/+)([^/]*)$"s... | Random trivia: drive letters aren't ascii, but are truncated utf16 (https://www.ryanliptak.com/blog/windows-drive-letters-are-not-limited-to-a-z/) | homedir,
isabspath,
isdirpath,
joinpath,
normpath,
realpath,
relpath,
splitdir,
splitdrive,
splitext,
splitpath
if Sys.isunix()
const path_separator = "/"
const path_separator_re = r"/+"sa # May be used by some external packages
@inline isseparator(c::Char) = ... | @@ -20,40 +22,128 @@ export
if Sys.isunix()
const path_separator = "/"
- const path_separator_re = r"/+"sa
- const path_directory_re = r"(?:^|/)\.{0,2}$"sa
- const path_dir_splitter = r"^(.*?)(/+)([^/]*)$"sa
- const path_ext_splitter = r"^((?:.*/)?(?:\.|[^/\.])[^/]*?)(\.[^/\.]*|)$"sa
+ const p... | base/path.jl | 26 | Julia | 0.357 | suggestion | 146 | 51 | 51 | false | Remove `Regex` / libpcre dependency from basic `path.jl` operations | 60,677 | JuliaLang/julia | 35,573 | Julia | vtjnash | johroj | Review this pull request change.
Language: Julia
File: base/path.jl
PR Title: Remove `Regex` / libpcre dependency from basic `path.jl` operations
BEFORE CODE: isdirpath,
joinpath,
normpath,
realpath,
relpath,
splitdir,
splitdrive,
splitext,
splitpath
if Sys.isunix()
const pa... | Random trivia: drive letters aren't ascii, but are truncated utf16 (https://www.ryanliptak.com/blog/windows-drive-letters-are-not-limited-to-a-z/) | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
ctx->ContextFlags = CONTEXT_CONTROL | CONTEXT_INTEGER;
if (!GetThreadContext(hThread, ctx)) {
if ((DWORD)-1 == ResumeThread(hThread))
abort();
return 0;
}
return 1;
}
void jl_thread_resume(int tid)
{
jl_ptls_t ptls2 = jl_atomic_load_relaxed(&jl_all_tls_states)[tid];
... | ```suggestion
int success = jl_thread_suspend_and_get_state(tid, 0, ctx);
if (jl_atomic_load_relaxed(&dll_notify_queue)) {
// if the dll_notify_queue is set, we lost the race against new code being added and cannot safely examine ctx
// resume the thread and let the outer loop retry (better ... | ctx->ContextFlags = CONTEXT_CONTROL | CONTEXT_INTEGER;
if (!GetThreadContext(hThread, ctx)) {
if ((DWORD)-1 == ResumeThread(hThread))
abort();
return 0;
}
return 1;
}
void jl_thread_resume(int tid)
{
jl_ptls_t ptls2 = jl_atomic_load_relaxed(&jl_all_tls_states)[tid];
... | @@ -475,6 +468,7 @@ int jl_thread_suspend(int16_t tid, bt_context_t *ctx)
{
jl_lock_profile(); // prevent concurrent mutation
uv_mutex_lock(&jl_in_stackwalk); // prevent multi-threaded dbghelp calls
+ jl_profile_process_dll_events(); // process any deferred dll events delayed due to lock contention
i... | src/signals-win.c | 26 | C | 1 | suggestion | 430 | 51 | 51 | false | [Windows] Add DLL notification queue to avoid deadlock | 60,690 | JuliaLang/julia | 35,573 | Julia | vtjnash | topolarity | Review this pull request change.
Language: C
File: src/signals-win.c
PR Title: [Windows] Add DLL notification queue to avoid deadlock
BEFORE CODE: ctx->ContextFlags = CONTEXT_CONTROL | CONTEXT_INTEGER;
if (!GetThreadContext(hThread, ctx)) {
if ((DWORD)-1 == ResumeThread(hThread))
abort()... | ```suggestion
int success = jl_thread_suspend_and_get_state(tid, 0, ctx);
if (jl_atomic_load_relaxed(&dll_notify_queue)) {
// if the dll_notify_queue is set, we lost the race against new code being added and cannot safely examine ctx
// resume the thread and let the outer loop retry (better ... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
event->DllBase = (uintptr_t)NotificationData->Unloaded.DllBase;
}
event->NotificationReason = NotificationReason;
event->next = jl_atomic_exchange_release(&dll_notify_queue, event);
if (uv_mutex_trylock(&jl_in_stackwalk) != 0)
return; // lock unavailable, process later
jl_profile_... | This is technically now a data race, since we requested that the profile thread concurrently read this object while we just got notified that it is about to be unmapped. Is there a way we can safely stall here (inside the close library lock) to cause profiling to stop looking at the earlier entry? |
dll_notification_event_t *next = event->next;
free(event);
event = next;
}
dll_notify_queue = NULL;
}
// Callback for LdrRegisterDllNotification
static VOID CALLBACK dll_notification_callback(
ULONG NotificationReason,
PCLDR_DLL_NOTIFICATION_DATA NotificationData,
PVOID Con... | @@ -503,29 +513,59 @@ typedef VOID CALLBACK (*PLDR_DLL_NOTIFICATION_FUNCTION)(
NTSTATUS NTAPI LdrRegisterDllNotification(ULONG Flags, PLDR_DLL_NOTIFICATION_FUNCTION NotificationFunction, PVOID Context, PVOID *Cookie);
NTSTATUS NTAPI LdrUnregisterDllNotification(PVOID Cookie);
+// caller should hold jl_in_stackwalk ... | src/stackwalk.c | 26 | C | 0.643 | bug | 298 | 51 | 51 | false | [Windows] Add DLL notification queue to avoid deadlock | 60,690 | JuliaLang/julia | 35,573 | Julia | vtjnash | topolarity | Review this pull request change.
Language: C
File: src/stackwalk.c
PR Title: [Windows] Add DLL notification queue to avoid deadlock
BEFORE CODE: event->DllBase = (uintptr_t)NotificationData->Unloaded.DllBase;
}
event->NotificationReason = NotificationReason;
event->next = jl_atomic_exchange_rel... | This is technically now a data race, since we requested that the profile thread concurrently read this object while we just got notified that it is about to be unmapped. Is there a way we can safely stall here (inside the close library lock) to cause profiling to stop looking at the earlier entry? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
event = next;
}
}
// Callback for LdrRegisterDllNotification
static VOID CALLBACK dll_notification_callback(
ULONG NotificationReason,
PCLDR_DLL_NOTIFICATION_DATA NotificationData,
PVOID Context)
{
(void)Context;
dll_notification_event_t *event = (dll_notification_event_t*)malloc(sizeo... | this releases event before storing to next, which makes any access (read/write) of next in the future UB | int prior_enqueued_load_of_same_module = 0;
dll_notification_event_t *prior_event = event;
while (prior_event->next) {
if (prior_event->next->DllBase == event->DllBase) {
assert(prior_event->next->NotificationReason == LDR_DLL_NOTIFICATION_REASON_L... | @@ -503,29 +513,59 @@ typedef VOID CALLBACK (*PLDR_DLL_NOTIFICATION_FUNCTION)(
NTSTATUS NTAPI LdrRegisterDllNotification(ULONG Flags, PLDR_DLL_NOTIFICATION_FUNCTION NotificationFunction, PVOID Context, PVOID *Cookie);
NTSTATUS NTAPI LdrUnregisterDllNotification(PVOID Cookie);
+// caller should hold jl_in_stackwalk ... | src/stackwalk.c | 26 | C | 0.357 | suggestion | 104 | 51 | 51 | false | [Windows] Add DLL notification queue to avoid deadlock | 60,690 | JuliaLang/julia | 35,573 | Julia | vtjnash | topolarity | Review this pull request change.
Language: C
File: src/stackwalk.c
PR Title: [Windows] Add DLL notification queue to avoid deadlock
BEFORE CODE: event = next;
}
}
// Callback for LdrRegisterDllNotification
static VOID CALLBACK dll_notification_callback(
ULONG NotificationReason,
PCLDR_DLL_NOTIF... | this releases event before storing to next, which makes any access (read/write) of next in the future UB | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
if (NotificationReason == LDR_DLL_NOTIFICATION_REASON_LOADED) {
const LDR_DLL_LOADED_NOTIFICATION_DATA *data = &NotificationData->Loaded;
event->FullDllName = _wcsdup(data->FullDllName->Buffer);
event->BaseDllName = _wcsdup(data->BaseDllName->Buffer);
event->DllBase = (uintptr_t)data... | It feels overly complicated to me for these events to get processed in two completely different places. And notably this one seems like wasted work since it is unreliable
```suggestion
``` | if (NotificationReason == LDR_DLL_NOTIFICATION_REASON_LOADED) {
const LDR_DLL_LOADED_NOTIFICATION_DATA *data = &NotificationData->Loaded;
event->FullDllName = _wcsdup(data->FullDllName->Buffer);
event->BaseDllName = _wcsdup(data->BaseDllName->Buffer);
event->DllBase = (uintptr_t)data... | @@ -503,36 +514,101 @@ typedef VOID CALLBACK (*PLDR_DLL_NOTIFICATION_FUNCTION)(
NTSTATUS NTAPI LdrRegisterDllNotification(ULONG Flags, PLDR_DLL_NOTIFICATION_FUNCTION NotificationFunction, PVOID Context, PVOID *Cookie);
NTSTATUS NTAPI LdrUnregisterDllNotification(PVOID Cookie);
+// caller should hold jl_in_stackwalk... | src/stackwalk.c | 26 | C | 0.714 | suggestion | 190 | 51 | 51 | false | [Windows] Add DLL notification queue to avoid deadlock | 60,690 | JuliaLang/julia | 35,573 | Julia | vtjnash | topolarity | Review this pull request change.
Language: C
File: src/stackwalk.c
PR Title: [Windows] Add DLL notification queue to avoid deadlock
BEFORE CODE: if (NotificationReason == LDR_DLL_NOTIFICATION_REASON_LOADED) {
const LDR_DLL_LOADED_NOTIFICATION_DATA *data = &NotificationData->Loaded;
event->FullDl... | It feels overly complicated to me for these events to get processed in two completely different places. And notably this one seems like wasted work since it is unreliable
```suggestion
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
memset(ctx, 0, sizeof(CONTEXT));
ctx->ContextFlags = CONTEXT_CONTROL | CONTEXT_INTEGER;
if (!GetThreadContext(hThread, ctx)) {
if ((DWORD)-1 == ResumeThread(hThread))
abort();
return 0;
}
return 1;
}
void jl_thread_resume(int tid)
{
jl_ptls_t ptls2 = jl_atomic_load_r... | You're missing a call to this function before any same-thread unwind (since it is now unreliable whether the trylock happens before you started running that code). Anywhere we took the jl_in_stackwalk lock (except when the victim thread is stopped) we need to now run this again (just like the old need_syminit flag appr... | memset(ctx, 0, sizeof(CONTEXT));
ctx->ContextFlags = CONTEXT_CONTROL | CONTEXT_INTEGER;
if (!GetThreadContext(hThread, ctx)) {
if ((DWORD)-1 == ResumeThread(hThread))
abort();
return 0;
}
return 1;
}
void jl_thread_resume(int tid)
{
jl_ptls_t ptls2 = jl_atomic_load_r... | @@ -475,7 +468,10 @@ int jl_thread_suspend(int16_t tid, bt_context_t *ctx)
{
jl_lock_profile(); // prevent concurrent mutation
uv_mutex_lock(&jl_in_stackwalk); // prevent multi-threaded dbghelp calls
+ uv_mutex_lock(&jl_dll_notify_lock);
+ jl_profile_process_dll_events(); // process any deferred dll ev... | src/signals-win.c | 26 | C | 0.571 | nitpick | 414 | 51 | 51 | false | [Windows] Add DLL notification queue to avoid deadlock | 60,690 | JuliaLang/julia | 35,573 | Julia | vtjnash | topolarity | Review this pull request change.
Language: C
File: src/signals-win.c
PR Title: [Windows] Add DLL notification queue to avoid deadlock
BEFORE CODE: memset(ctx, 0, sizeof(CONTEXT));
ctx->ContextFlags = CONTEXT_CONTROL | CONTEXT_INTEGER;
if (!GetThreadContext(hThread, ctx)) {
if ((DWORD)-1 == Resum... | You're missing a call to this function before any same-thread unwind (since it is now unreliable whether the trylock happens before you started running that code). Anywhere we took the jl_in_stackwalk lock (except when the victim thread is stopped) we need to now run this again (just like the old need_syminit flag appr... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
AA.registerFunctionAnalysis<ScopedNoAliasAA>();
AA.registerFunctionAnalysis<TypeBasedAA>();
}
TM.registerDefaultAliasAnalyses(AA);
return AA;
});
// Register our TargetLibraryInfoImpl.
FAM.registerPass([&] JL_NOTSAFEPOINT { retu... | Please use a real parser (even if it is just getopt) and not this claude garbage. If we want to act like LLVM, it should use the function structure here: https://llvm.org/doxygen/CommandLine_8cpp_source.html#l01462
e.g. tokenization by [TokenizeGNUCommandLine](https://llvm.org/doxygen/namespacellvm_1_1cl.html#aff28c... | AA.registerFunctionAnalysis<BasicAA>();
AA.registerFunctionAnalysis<ScopedNoAliasAA>();
AA.registerFunctionAnalysis<TypeBasedAA>();
}
TM.registerDefaultAliasAnalyses(AA);
return AA;
});
// Register our TargetLibraryInfoI... | @@ -738,8 +740,55 @@ PIC.addClassToPassName(decltype(CREATE_PASS)::name(), NAME);
}
}
-NewPM::NewPM(std::unique_ptr<TargetMachine> TM, OptimizationLevel O, OptimizationOptions options) :
- TM(std::move(TM)), O(O), options(options), TimePasses() {}
+// Parse LLVM-style option string into PrintOptions
+void pa... | src/pipeline.cpp | 26 | C++ | 0.5 | suggestion | 676 | 51 | 51 | false | codegen: Add llvm_options parameter to code_llvm for pass instrumentation | 60,698 | JuliaLang/julia | 35,573 | Julia | vtjnash | gbaraldi | Review this pull request change.
Language: C++
File: src/pipeline.cpp
PR Title: codegen: Add llvm_options parameter to code_llvm for pass instrumentation
BEFORE CODE: AA.registerFunctionAnalysis<ScopedNoAliasAA>();
AA.registerFunctionAnalysis<TypeBasedAA>();
}
... | Please use a real parser (even if it is just getopt) and not this claude garbage. If we want to act like LLVM, it should use the function structure here: https://llvm.org/doxygen/CommandLine_8cpp_source.html#l01462
e.g. tokenization by [TokenizeGNUCommandLine](https://llvm.org/doxygen/namespacellvm_1_1cl.html#aff28c... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
llvmf_dump = Ref{LLVMFDump}()
@ccall jl_get_llvmf_defn(llvmf_dump::Ptr{LLVMFDump}, mi::Any, src::Any, wrapper::Bool,
true::Bool, ""::Cstring, params::CodegenParams)::Cvoid
llvmf_dump[].f == C_NULL && error("could not compile the specified method")
str = @ccall jl_dump_functi... | This should be freed by the same shared library that allocated it. Freeing objects from a different place than you allocated it leads to awkwardness (aka crashes) when you try to use alternatively allocators like mimalloc, and is an anti-pattern that should be always avoided | strip_ir_metadata::Bool, dump_module::Bool,
optimize::Bool, debuginfo::Symbol,
llvm_options::String,
params::CodegenParams)
llvmf_dump = Ref{LLVMFDump}()
@ccall jl_get_llvmf_defn(llvmf_dump::Ptr{LLVMFDump}, mi::Any, src::Any,
wrapper::Bool, optimize::... | @@ -312,18 +314,30 @@ function _dump_function_llvm(
mi::Core.MethodInstance, src::Core.CodeInfo, wrapper::Bool,
strip_ir_metadata::Bool, dump_module::Bool,
optimize::Bool, debuginfo::Symbol,
+ llvm_options::String,
params::CodegenParams)
llvmf_dump = Ref{LLVMFDump}()
... | stdlib/InteractiveUtils/src/codeview.jl | 26 | Julia | 0.5 | bug | 275 | 51 | 51 | false | codegen: Add llvm_options parameter to code_llvm for pass instrumentation | 60,698 | JuliaLang/julia | 35,573 | Julia | vtjnash | gbaraldi | Review this pull request change.
Language: Julia
File: stdlib/InteractiveUtils/src/codeview.jl
PR Title: codegen: Add llvm_options parameter to code_llvm for pass instrumentation
BEFORE CODE: llvmf_dump = Ref{LLVMFDump}()
@ccall jl_get_llvmf_defn(llvmf_dump::Ptr{LLVMFDump}, mi::Any, src::Any, wrapper::Bool,... | This should be freed by the same shared library that allocated it. Freeing objects from a different place than you allocated it leads to awkwardness (aka crashes) when you try to use alternatively allocators like mimalloc, and is an anti-pattern that should be always avoided | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
end
function _dump_function_llvm(
mi::Core.MethodInstance, src::Core.CodeInfo, wrapper::Bool,
strip_ir_metadata::Bool, dump_module::Bool,
optimize::Bool, debuginfo::Symbol,
llvm_options::String,
params::CodegenParams)
llvmf_dump = Ref{LLVMFDump}()
@ccall jl_get_llvmf_def... | I makes me wonder if we want to support more (all) of the opt pipeline?
In particular remarks? But perhaps custom opt pipeline would also make sense | llvmf_dump[].f == C_NULL && error("could not compile the specified method")
# jl_dump_function_ir handles pass_output internally (prepends it and frees it)
str = @ccall jl_dump_function_ir(llvmf_dump::Ptr{LLVMFDump}, strip_ir_metadata::Bool,
dump_module::Bool, debuginfo:... | @@ -333,12 +347,23 @@ All metadata and dbg.* calls are removed from the printed bitcode. For the full
To dump the entire module that encapsulates the function (with declarations), set the `dump_module` keyword to true.
Keyword argument `debuginfo` may be one of source (default) or none, to specify the verbosity of co... | stdlib/InteractiveUtils/src/codeview.jl | 26 | Julia | 0.357 | suggestion | 149 | 51 | 51 | false | codegen: Add llvm_options parameter to code_llvm for pass instrumentation | 60,698 | JuliaLang/julia | 35,573 | Julia | vchuravy | gbaraldi | Review this pull request change.
Language: Julia
File: stdlib/InteractiveUtils/src/codeview.jl
PR Title: codegen: Add llvm_options parameter to code_llvm for pass instrumentation
BEFORE CODE: end
function _dump_function_llvm(
mi::Core.MethodInstance, src::Core.CodeInfo, wrapper::Bool,
strip_ir_meta... | I makes me wonder if we want to support more (all) of the opt pipeline?
In particular remarks? But perhaps custom opt pipeline would also make sense | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
// Parse LLVM-style option string into PrintOptions using LLVM's tokenizer
void parseLLVMOptions(const char *options, PrintOptions &out) JL_NOTSAFEPOINT {
if (!options || options[0] == '\0')
return;
// Tokenize the options string using LLVM's GNU command line tokenizer
BumpPtrAllocator Alloc;
... | ```suggestion
} else if (Arg.starts_with("-print-after=")) {
out.print_after = Arg.substr(13).str();
} else if (Arg == "-print-after") {
out.print_after = std::string(Argv[++i]);
```
and so on, with bounds checking of course |
// Parse LLVM-style option string into PrintOptions using LLVM's tokenizer
void parseLLVMOptions(const char *options, PrintOptions &out) JL_NOTSAFEPOINT {
if (!options || options[0] == '\0')
return;
// Tokenize the options string using LLVM's GNU command line tokenizer
BumpPtrAllocator Alloc;
... | @@ -738,8 +741,43 @@ PIC.addClassToPassName(decltype(CREATE_PASS)::name(), NAME);
}
}
-NewPM::NewPM(std::unique_ptr<TargetMachine> TM, OptimizationLevel O, OptimizationOptions options) :
- TM(std::move(TM)), O(O), options(options), TimePasses() {}
+// Parse LLVM-style option string into PrintOptions using LL... | src/pipeline.cpp | 26 | C++ | 0.786 | suggestion | 270 | 51 | 51 | false | codegen: Add llvm_options parameter to code_llvm for pass instrumentation | 60,698 | JuliaLang/julia | 35,573 | Julia | vtjnash | gbaraldi | Review this pull request change.
Language: C++
File: src/pipeline.cpp
PR Title: codegen: Add llvm_options parameter to code_llvm for pass instrumentation
BEFORE CODE:
// Parse LLVM-style option string into PrintOptions using LLVM's tokenizer
void parseLLVMOptions(const char *options, PrintOptions &out) JL_NOTSAFEP... | ```suggestion
} else if (Arg.starts_with("-print-after=")) {
out.print_after = Arg.substr(13).str();
} else if (Arg == "-print-after") {
out.print_after = std::string(Argv[++i]);
```
and so on, with bounds checking of course | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
return str
end
function _dump_function_llvm(
mi::Core.MethodInstance, src::Core.CodeInfo, wrapper::Bool,
strip_ir_metadata::Bool, dump_module::Bool,
optimize::Bool, debuginfo::Symbol,
llvm_options::String,
params::CodegenParams)
llvmf_dump = Ref{LLVMFDump}()
@ccall j... | Should we prepend `;` to each newline in the annotations stream? I think normally in LLVM it might go to stderr while the final output goes to stdout, so they are normally separated that way instead?
```julia
str = replace(str, '\n' => "\n;")
``` | params::CodegenParams)::Cvoid
llvmf_dump[].f == C_NULL && error("could not compile the specified method")
# jl_dump_function_ir handles pass_output internally (prepends it and frees it)
str = @ccall jl_dump_function_ir(llvmf_dump::Ptr{LLVMFDump}, strip_ir_metadata::Bool,
... | @@ -312,18 +314,30 @@ function _dump_function_llvm(
mi::Core.MethodInstance, src::Core.CodeInfo, wrapper::Bool,
strip_ir_metadata::Bool, dump_module::Bool,
optimize::Bool, debuginfo::Symbol,
+ llvm_options::String,
params::CodegenParams)
llvmf_dump = Ref{LLVMFDump}()
... | stdlib/InteractiveUtils/src/codeview.jl | 26 | Julia | 0.857 | question | 249 | 51 | 51 | false | codegen: Add llvm_options parameter to code_llvm for pass instrumentation | 60,698 | JuliaLang/julia | 35,573 | Julia | vtjnash | gbaraldi | Review this pull request change.
Language: Julia
File: stdlib/InteractiveUtils/src/codeview.jl
PR Title: codegen: Add llvm_options parameter to code_llvm for pass instrumentation
BEFORE CODE: return str
end
function _dump_function_llvm(
mi::Core.MethodInstance, src::Core.CodeInfo, wrapper::Bool,
... | Should we prepend `;` to each newline in the annotations stream? I think normally in LLVM it might go to stderr while the final output goes to stdout, so they are normally separated that way instead?
```julia
str = replace(str, '\n' => "\n;")
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
metadataM.setOverrideStackAlignment(OverrideStackAlignment);
// reflect the address of the jl_RTLD_DEFAULT_handle variable
// back to the caller, so that we can check for consistency issues
GlobalValue *jlRTLD_DEFAULT_var = jl_emit_RTLD_DEFAULT_var(&metadataM);
Type *T_size = D... | Looks like jl_dump_function_asm isn't freeing the newly added memory? | jl_printf(stream, "\n---- optimized IR ----\n");
jl_get_llvmf_defn(&llvmf_dump, mi, src, 0, true, nullptr, jl_default_cgparams);
if (llvmf_dump.F) {
jl_value_t *ir = jl_dump_function_ir(&llvmf_dump, 0, 1, "source");
if (ir != NULL && jl_is_string(ir))
jl_printf(stream, "%s", jl_s... | @@ -2552,11 +2552,12 @@ extern "C" JL_DLLEXPORT_CODEGEN jl_code_info_t *jl_gdbdumpcode(jl_method_instanc
// for use in reflection from Julia.
// This is paired with jl_dump_function_ir and jl_dump_function_asm, either of which will free all memory allocated here | src/aotcompile.cpp | 26 | C++ | 0.357 | question | 69 | 51 | 51 | false | codegen: Add llvm_options parameter to code_llvm for pass instrumentation | 60,698 | JuliaLang/julia | 35,573 | Julia | vtjnash | gbaraldi | Review this pull request change.
Language: C++
File: src/aotcompile.cpp
PR Title: codegen: Add llvm_options parameter to code_llvm for pass instrumentation
BEFORE CODE: metadataM.setOverrideStackAlignment(OverrideStackAlignment);
// reflect the address of the jl_RTLD_DEFAULT_handle variable
... | Looks like jl_dump_function_asm isn't freeing the newly added memory? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
// Check for "=" syntax first
if (Arg.size() > prefix.size() && Arg[prefix.size()] == '=') {
return Arg.substr(prefix.size() + 1);
}
// Otherwise, consume the next argument
if (idx + 1 < Argv.size()) {
return StringRef(Argv[++idx]);
}
errs(... | Can you ask Claude to rewrite this so that is isn't doing just a prefix check, but can have `getNextValue` implement the exact checks:
```suggestion
} else if (StringRef val = getNextValue(i, Arg, "-print-after=")) {
```
If you were going to do a prefix comparison scan only, the getopt standard requires ... | // Supports both "-option=value" and "-option value" syntax
auto getNextValue = [&](size_t &idx, StringRef Arg, StringRef option) JL_NOTSAFEPOINT -> StringRef {
StringRef optionName = option.drop_back(); // remove trailing "="
// Check for "-option=value" syntax
if (Arg.starts_with(optio... | @@ -738,8 +741,64 @@ PIC.addClassToPassName(decltype(CREATE_PASS)::name(), NAME);
}
}
-NewPM::NewPM(std::unique_ptr<TargetMachine> TM, OptimizationLevel O, OptimizationOptions options) :
- TM(std::move(TM)), O(O), options(options), TimePasses() {}
+// Parse LLVM-style option string into PrintOptions using LL... | src/pipeline.cpp | 26 | C++ | 0.786 | suggestion | 568 | 51 | 51 | false | codegen: Add llvm_options parameter to code_llvm for pass instrumentation | 60,698 | JuliaLang/julia | 35,573 | Julia | vtjnash | gbaraldi | Review this pull request change.
Language: C++
File: src/pipeline.cpp
PR Title: codegen: Add llvm_options parameter to code_llvm for pass instrumentation
BEFORE CODE: // Check for "=" syntax first
if (Arg.size() > prefix.size() && Arg[prefix.size()] == '=') {
return Arg.substr(prefix.siz... | Can you ask Claude to rewrite this so that is isn't doing just a prefix check, but can have `getNextValue` implement the exact checks:
```suggestion
} else if (StringRef val = getNextValue(i, Arg, "-print-after=")) {
```
If you were going to do a prefix comparison scan only, the getopt standard requires ... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
global.second->setLinkage(GlobalValue::PrivateLinkage);
global.second->setUnnamedAddr(GlobalValue::UnnamedAddr::Global);
global.second->setVisibility(GlobalValue::DefaultVisibility);
}
}
if (!jl_o... | jl_free_llvmf_pass_output doesn't exist | // This file is a part of Julia. License is MIT: https://julialang.org/license
#include "llvm-version.h"
#include "platform.h"
// target support
#include <llvm/TargetParser/Triple.h>
#include "llvm/Support/CodeGen.h"
#include <llvm/ADT/Statistic.h>
#include <llvm/Analysis/TargetLibraryInfo.h>
#include <llvm/Analysis/... | @@ -2645,10 +2646,21 @@ void jl_get_llvmf_defn_impl(jl_llvmf_dump_t *dump, jl_method_instance_t *mi, jl_
opts.sanitize_memory = params.sanitize_memory;
opts.sanitize_thread = params.sanitize_thread;
opts.sanitize_address = params.sanitize_address;
- ... | src/aotcompile.cpp | 26 | C++ | 0.071 | suggestion | 39 | 51 | 2,499 | false | codegen: Add llvm_options parameter to code_llvm for pass instrumentation | 60,698 | JuliaLang/julia | 35,573 | Julia | vtjnash | gbaraldi | Review this pull request change.
Language: C++
File: src/aotcompile.cpp
PR Title: codegen: Add llvm_options parameter to code_llvm for pass instrumentation
BEFORE CODE: global.second->setLinkage(GlobalValue::PrivateLinkage);
global.second->setUnnamedAddr(GlobalValue::... | jl_free_llvmf_pass_output doesn't exist | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
bool sanitize_thread=true,
#else
bool sanitize_thread=false,
#endif
#ifdef _COMPILER_ASAN_ENABLED_
bool sanitize_address=true
#else
bool sanitize_address=false
#endif
) JL_NOTSAFEPOINT {
return {lower_intrinsics, dump_native, external_use, llvm_only,
always_inline... | That isn't a very useful link. The link for the ChangeReporter is here:
https://llvm.org/doxygen/classllvm_1_1ChangeReporter.html#a7a57fcd8f955992d5167dae6978748f2
(and is simply a text diff, nothing fancy, no "integration" is required) | bool sanitize_thread=true,
#else
bool sanitize_thread=false,
#endif
#ifdef _COMPILER_ASAN_ENABLED_
bool sanitize_address=true
#else
bool sanitize_address=false
#endif
) JL_NOTSAFEPOINT {
return {lower_intrinsics, dump_native, external_use, llvm_only,
always_inline... | @@ -132,12 +132,42 @@ struct OptimizationOptions {
}
};
+struct PrintOptions {
+ bool print_before_all = false;
+ bool print_after_all = false;
+ bool print_module_scope = false;
+ // TODO: Add print_changed support. This would require integrating with LLVM's
+ // ChangedIRPrinter infrastructure ... | src/jitlayers.h | 26 | C/C++ | 0.5 | suggestion | 237 | 51 | 51 | false | codegen: Add llvm_options parameter to code_llvm for pass instrumentation | 60,698 | JuliaLang/julia | 35,573 | Julia | vtjnash | gbaraldi | Review this pull request change.
Language: C/C++
File: src/jitlayers.h
PR Title: codegen: Add llvm_options parameter to code_llvm for pass instrumentation
BEFORE CODE: bool sanitize_thread=true,
#else
bool sanitize_thread=false,
#endif
#ifdef _COMPILER_ASAN_ENABLED_
bool sanitize_address=tru... | That isn't a very useful link. The link for the ChangeReporter is here:
https://llvm.org/doxygen/classllvm_1_1ChangeReporter.html#a7a57fcd8f955992d5167dae6978748f2
(and is simply a text diff, nothing fancy, no "integration" is required) | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
void parseLLVMOptions(const char *options, PrintOptions &out) JL_NOTSAFEPOINT {
if (!options || options[0] == '\0')
return;
// Tokenize the options string using LLVM's GNU command line tokenizer
BumpPtrAllocator Alloc;
StringSaver Saver(Alloc);
SmallVector<const char *, 16> Argv;
cl::To... | Can you add the error string to PrintOptions, to print in the output, rather than relying on the `errs()` stream working correctly here | void parseLLVMOptions(const char *options, PrintOptions &out) JL_NOTSAFEPOINT {
if (!options || options[0] == '\0')
return;
// Tokenize the options string using LLVM's GNU command line tokenizer
BumpPtrAllocator Alloc;
StringSaver Saver(Alloc);
SmallVector<const char *, 16> Argv;
cl::To... | @@ -738,8 +741,61 @@ PIC.addClassToPassName(decltype(CREATE_PASS)::name(), NAME);
}
}
-NewPM::NewPM(std::unique_ptr<TargetMachine> TM, OptimizationLevel O, OptimizationOptions options) :
- TM(std::move(TM)), O(O), options(options), TimePasses() {}
+// Parse LLVM-style option string into PrintOptions using LL... | src/pipeline.cpp | 26 | C++ | 0.643 | suggestion | 135 | 51 | 51 | false | codegen: Add llvm_options parameter to code_llvm for pass instrumentation | 60,698 | JuliaLang/julia | 35,573 | Julia | vtjnash | gbaraldi | Review this pull request change.
Language: C++
File: src/pipeline.cpp
PR Title: codegen: Add llvm_options parameter to code_llvm for pass instrumentation
BEFORE CODE: void parseLLVMOptions(const char *options, PrintOptions &out) JL_NOTSAFEPOINT {
if (!options || options[0] == '\0')
return;
// Token... | Can you add the error string to PrintOptions, to print in the output, rather than relying on the `errs()` stream working correctly here | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
//We must recreate the analysis managers every time
//so that analyses from previous runs of the pass manager
//do not hang around for the next run
StandardInstrumentations SI(M.getContext(), /*DebugLogging=*/false);
PassInstrumentationCallbacks PIC;
adjustPIC(PIC);
TimePasses.registerCallba... | This should be a comma-separated list (e.g. use split in a loop when parsing, since the user also can pass this flag multiple times). See https://llvm.org/doxygen/CommandLine_8cpp.html#a030478956c379ec2bb71550cb0526fb6 (CommaSeparateAndAddOccurrence) | } else if (const auto *SCC = unwrapIR<LazyCallGraph::SCC>(IR)) {
for (auto &CGN : *SCC) {
Function &F = CGN.getFunction();
F.print(OS);
}
} else {
OS << "Unknown IR type\n";
}
}
void NewPM::run(Module &M) {
//We must recreate the analysis managers every t... | @@ -762,14 +818,97 @@ AnalysisManagers::AnalysisManagers(PassBuilder &PB) : LAM(), FAM(), CGAM(), MAM(
AnalysisManagers::~AnalysisManagers() = default;
+// Helper to unwrap IR from Any to a specific type
+template <typename IRType>
+static const IRType *unwrapIR(Any IR) JL_NOTSAFEPOINT {
+ const IRType *const *... | src/pipeline.cpp | 26 | C++ | 0.5 | suggestion | 250 | 51 | 51 | false | codegen: Add llvm_options parameter to code_llvm for pass instrumentation | 60,698 | JuliaLang/julia | 35,573 | Julia | vtjnash | gbaraldi | Review this pull request change.
Language: C++
File: src/pipeline.cpp
PR Title: codegen: Add llvm_options parameter to code_llvm for pass instrumentation
BEFORE CODE: //We must recreate the analysis managers every time
//so that analyses from previous runs of the pass manager
//do not hang around for th... | This should be a comma-separated list (e.g. use split in a loop when parsing, since the user also can pass this flag multiple times). See https://llvm.org/doxygen/CommandLine_8cpp.html#a030478956c379ec2bb71550cb0526fb6 (CommaSeparateAndAddOccurrence) | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
}
void Optimizer::initializeAlloca(IRBuilder<> &prolog_builder, AllocaInst *buff, AllocFnKind allockind)
{
if ((allockind & AllocFnKind::Uninitialized) != AllocFnKind::Unknown)
return;
assert(!buff->isArrayAllocation());
Type *T = buff->getAllocatedType();
const DataLayout &DL = F.getParent()->... | This feels loosely unsound, since this is the minimum known alignment, and not the required alignment. The JL_SMALL_BYTE_ALIGNMENT value is the largest value that julia.gc_alloc_obj is permitted to return, so it is sometimes reasonable that we can use this as a hint, but we should be sure to clarify that this overalign... | }
void Optimizer::initializeAlloca(IRBuilder<> &prolog_builder, AllocaInst *buff, AllocFnKind allockind)
{
if ((allockind & AllocFnKind::Uninitialized) != AllocFnKind::Unknown)
return;
assert(!buff->isArrayAllocation());
Type *T = buff->getAllocatedType();
const DataLayout &DL = F.getParent()->... | @@ -678,11 +678,8 @@ void Optimizer::moveToStack(CallInst *orig_inst, size_t sz, bool has_ref, AllocF
// The allocation does not escape or get used in a phi node so none of the derived
// SSA from it are live when we run the allocation again.
// It is now safe to promote the allocation to an entry block ... | src/llvm-alloc-opt.cpp | 26 | C++ | 1 | performance | 543 | 51 | 51 | false | llvm: Fix alloca alignment and type selection in AllocOpt | 60,699 | JuliaLang/julia | 35,573 | Julia | vtjnash | gbaraldi | Review this pull request change.
Language: C++
File: src/llvm-alloc-opt.cpp
PR Title: llvm: Fix alloca alignment and type selection in AllocOpt
BEFORE CODE: }
void Optimizer::initializeAlloca(IRBuilder<> &prolog_builder, AllocaInst *buff, AllocFnKind allockind)
{
if ((allockind & AllocFnKind::Uninitialized) !=... | This feels loosely unsound, since this is the minimum known alignment, and not the required alignment. The JL_SMALL_BYTE_ALIGNMENT value is the largest value that julia.gc_alloc_obj is permitted to return, so it is sometimes reasonable that we can use this as a hint, but we should be sure to clarify that this overalign... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
kind(rhs) in KSet"new splatnew cfunction isdefined call foreigncall gc_preserve_begin foreigncall new_opaque_closure")
end
function check_no_local_bindings(ctx, ex, msg)
contains_nonglobal_binding = contains_unquoted(ex) do e
kind(e) == K"BindingId" && get_binding(ctx, e).kind !== :global
e... | ```suggestion
elseif ((all_simple || is_const_read_arg(ctx, arg_val)) &&
is_valid_body_ir_argument(ctx, arg_val))
```
Aligning to the inner block is confusing IMO.
Although I actually don't think we have a great style for breaking conditions on multiple lines like this. | kind(rhs) in KSet"new splatnew cfunction isdefined call foreigncall gc_preserve_begin foreigncall new_opaque_closure")
end
function check_no_local_bindings(ctx, ex, msg)
contains_nonglobal_binding = contains_unquoted(ex) do e
kind(e) == K"BindingId" && get_binding(ctx, e).kind !== :global
e... | @@ -161,7 +161,11 @@ function compile_args(ctx, args)
args_out = SyntaxList(ctx)
for arg in args
arg_val = compile(ctx, arg, true, false)
- if (all_simple || is_const_read_arg(ctx, arg_val)) && is_valid_body_ir_argument(ctx, arg_val)
+ if isnothing(arg_val)
+ # arguments that... | JuliaLowering/src/linear_ir.jl | 26 | Julia | 0.786 | suggestion | 303 | 51 | 51 | false | [JuliaLowering] Fix linearize step when arguments have no value | 60,704 | JuliaLang/julia | 35,573 | Julia | topolarity | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/linear_ir.jl
PR Title: [JuliaLowering] Fix linearize step when arguments have no value
BEFORE CODE: kind(rhs) in KSet"new splatnew cfunction isdefined call foreigncall gc_preserve_begin foreigncall new_opaque_closure")
end
functio... | ```suggestion
elseif ((all_simple || is_const_read_arg(ctx, arg_val)) &&
is_valid_body_ir_argument(ctx, arg_val))
```
Aligning to the inner block is confusing IMO.
Although I actually don't think we have a great style for breaking conditions on multiple lines like this. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# TODO: Should every form get layerid systematically? Or only the ones
# which expand_forms_2 needs?
layerid = get(ex, :scope_layer, current_layer_id(ctx))
setattr(mapchildren(e->expand_forms_1(ctx,e), ctx, ex),
:scope_layer, layerid)
else
mapchildren(e->expan... | What prevents this from accidentally transforming a quoted `if @generated ...`? Is it just that macro-expansion hasn't happened yet, so it's impossible to "type" `if $(Expr(:generated)) ...` ?
(side note) this pattern matching turned out quite nice, I can see why you wanted the utility 👍 | # TODO: Should every form get layerid systematically? Or only the ones
# which expand_forms_2 needs?
layerid = get(ex, :scope_layer, current_layer_id(ctx))
setattr(mapchildren(e->expand_forms_1(ctx,e), ctx, ex),
:scope_layer, layerid)
else
mapchildren(e->expan... | @@ -564,6 +572,20 @@ function expand_forms_1(ctx::MacroExpansionContext, ex::SyntaxTree)
end
end
+has_if_generated(st::SyntaxTree) = JuliaSyntax.@stm st begin
+ (_, when=is_leaf(st)||is_quoted(st)) -> false
+ [K"function" _...] -> false
+ ([K"=" call _], when=is_eventually_call(call)) -> false
+ [K"... | JuliaLowering/src/macro_expansion.jl | 26 | Julia | 0.714 | refactor | 292 | 50 | 51 | false | [JuliaLowering] Quote generated function body before desugaring | 60,710 | JuliaLang/julia | 35,573 | Julia | topolarity | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/macro_expansion.jl
PR Title: [JuliaLowering] Quote generated function body before desugaring
BEFORE CODE: # TODO: Should every form get layerid systematically? Or only the ones
# which expand_forms_2 needs?
layerid = ge... | What prevents this from accidentally transforming a quoted `if @generated ...`? Is it just that macro-expansion hasn't happened yet, so it's impossible to "type" `if $(Expr(:generated)) ...` ?
(side note) this pattern matching turned out quite nice, I can see why you wanted the utility 👍 | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
gen = expand_forms_1(ctx, expand_quote(
ctx, @ast ctx e2 [K"block" split_generated(e2, true)]))
nongen = split_generated(e2, false)
@ast ctx ex [K"generated_function" e1 gen nongen]
elseif k == K"<:" || k == K">:" || k == K"-->"
# TODO: Should every form get layerid syste... | nitpick, but this ternary is a bit hard to read
```suggestion
[K"if" [K"generated"] gen nongen] -> if gen_part
@ast st._graph st [K"$" gen]
else
nongen
end
``` | gen = expand_forms_1(ctx, expand_quote(
ctx, @ast ctx e2 [K"block" split_generated(e2, true)]))
nongen = split_generated(e2, false)
@ast ctx ex [K"generated_function" e1 gen nongen]
elseif k == K"<:" || k == K">:" || k == K"-->"
# TODO: Should every form get layerid syste... | @@ -564,6 +572,20 @@ function expand_forms_1(ctx::MacroExpansionContext, ex::SyntaxTree)
end
end
+has_if_generated(st::SyntaxTree) = JuliaSyntax.@stm st begin
+ (_, when=is_leaf(st)||is_quoted(st)) -> false
+ [K"function" _...] -> false
+ ([K"=" call _], when=is_eventually_call(call)) -> false
+ [K"... | JuliaLowering/src/macro_expansion.jl | 26 | Julia | 0.714 | suggestion | 194 | 51 | 51 | false | [JuliaLowering] Quote generated function body before desugaring | 60,710 | JuliaLang/julia | 35,573 | Julia | topolarity | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/macro_expansion.jl
PR Title: [JuliaLowering] Quote generated function body before desugaring
BEFORE CODE: gen = expand_forms_1(ctx, expand_quote(
ctx, @ast ctx e2 [K"block" split_generated(e2, true)]))
nongen = spli... | nitpick, but this ternary is a bit hard to read
```suggestion
[K"if" [K"generated"] gen nongen] -> if gen_part
@ast st._graph st [K"$" gen]
else
nongen
end
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# This file is a part of Julia. License is MIT: https://julialang.org/license
include("tempdepot.jl")
function precompile_test_harness(@nospecialize(f), testset::String)
@testset "$testset" precompile_test_harness(f, true)
end
function precompile_test_harness(@nospecialize(f), separate::Bool=true)
load_path =... | Something probably went horribly wrong if `ci.max_world != WORLD_AGE_REVALIDATION_SENTINEL`, since that means the Compiler made a mistake and forgot to validate it | try
pushfirst!(LOAD_PATH, load_path)
pushfirst!(DEPOT_PATH, load_cache_path)
f(load_path)
finally
filter!((≠)(load_path), LOAD_PATH)
separate && filter!((≠)(load_cache_path), DEPOT_PATH)
end
return nothing
end
let original_depot_path = copy(Base.DEPOT_PATH)
o... | @@ -29,3 +29,14 @@ let original_depot_path = copy(Base.DEPOT_PATH)
append!(Base.LOAD_PATH, original_load_path)
end
end
+
+function check_presence(mi, token)
+ ci = isdefined(mi, :cache) ? mi.cache : nothing
+ while ci !== nothing
+ if ci.owner === token && ci.max_world == typemax(UInt) | test/precompile_utils.jl | 26 | Julia | 0.5 | bug | 163 | 32 | 33 | false | [1.12] Don't skip CIs from external absints | 60,741 | JuliaLang/julia | 35,573 | Julia | vtjnash | gbaraldi | Review this pull request change.
Language: Julia
File: test/precompile_utils.jl
PR Title: [1.12] Don't skip CIs from external absints
BEFORE CODE: # This file is a part of Julia. License is MIT: https://julialang.org/license
include("tempdepot.jl")
function precompile_test_harness(@nospecialize(f), testset::Strin... | Something probably went horribly wrong if `ci.max_world != WORLD_AGE_REVALIDATION_SENTINEL`, since that means the Compiler made a mistake and forgot to validate it | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@test !success(cmd)
end
end
## `Main.main` entrypoint
# Basic usage
@test readchomp(`$(Base.julia_cmd()) -e '(@main)(args) = println("hello")'`) == "hello"
# Test ARGS with -e
@test readchomp(`$(Base.julia_cmd()) -e '(@main)(args) = println(args)' a b`) == repr(["a", "b"])
# Test import from module
@test readch... | This is kind of what `@testset let` is for, but that's a nit. | MAX32HEAP = 1536 * 1024 * 1024
if Int === Int32 && hint > MAX32HEAP
hint = MAX32HEAP
end
@test abs(Float64(maxmem) - hint)/maxmem < 0.05
@test readchomp(`$(Base.julia_cmd()) --startup-file=no --heap-size-hint=10M -e "println(@ccall jl_gc_get_max_memory()::UInt64)"`) == "$(1*1024*1024)"
end
... | @@ -1364,9 +1364,20 @@ end
# test --bug-report=rr
if Sys.islinux() && Sys.ARCH in (:i686, :x86_64) # rr is only available on these platforms
mktempdir() do temp_trace_dir
- test_read_success(setenv(`$(Base.julia_cmd()) --bug-report=rr-local -e 'exit()'`,
- "JULIA_RR_RECORD_... | test/cmdlineargs.jl | 26 | Julia | 0.357 | suggestion | 61 | 51 | 51 | false | Treat 'unable to open perf counter' as success in rr-related tests | 60,780 | JuliaLang/julia | 35,573 | Julia | Keno | qinsoon | Review this pull request change.
Language: Julia
File: test/cmdlineargs.jl
PR Title: Treat 'unable to open perf counter' as success in rr-related tests
BEFORE CODE: @test !success(cmd)
end
end
## `Main.main` entrypoint
# Basic usage
@test readchomp(`$(Base.julia_cmd()) -e '(@main)(args) = println("hello")'`) ... | This is kind of what `@testset let` is for, but that's a nit. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
record_reason(reasons, "different Julia build configuration")
return true # incompatible cache file
end
modules, (includes, _, requires), required_modules, srctextpos, prefs, prefs_hash, clone_targets, actual_flags, syntax_version = parse_cache_header(io, cachefile)
if is... | Is `stalecheck` really right here? This is more like the cache flag above. Isn't the issue that require_stdlib is trying to load with the wrong syntax version? | catch ex
ex isa IOError || ex isa SystemError || rethrow()
@debug "Rejecting cache file $cachefile for $modkey because it could not be opened" isfile(cachefile)
return true
end
try
checksum = isvalid_cache_header(io)
if iszero(checksum)
@debug "Rejecting c... | @@ -4286,7 +4286,7 @@ end
record_reason(reasons, "different compilation options")
return true
end
- if syntax_version != cache_syntax_version(modspec.julia_syntax_version)
+ if stalecheck && syntax_version != cache_syntax_version(modspec.julia_syntax_version) | base/loading.jl | 26 | Julia | 0.5 | bug | 159 | 51 | 51 | false | Skip syntax check in require_stdlib. Add test. Update stdlib manifest | 60,791 | JuliaLang/julia | 35,573 | Julia | Keno | IanButterworth | Review this pull request change.
Language: Julia
File: base/loading.jl
PR Title: Skip syntax check in require_stdlib. Add test. Update stdlib manifest
BEFORE CODE: record_reason(reasons, "different Julia build configuration")
return true # incompatible cache file
end
modules,... | Is `stalecheck` really right here? This is more like the cache flag above. Isn't the issue that require_stdlib is trying to load with the wrong syntax version? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@test invokelatest(getglobal, (@eval (using VersionedDep1; VersionedDep1)), :ver) == v"1.13"
# syntax.julia_version from Manifest = "1.14"
@test invokelatest(getglobal, (@eval (using VersionedDep2; VersionedDep2)), :ver) == v"1.14"
# syntax.julia_version from Manifest = "1.0" should be c... | Just realized this test needs to monitor stderr. This doesn't currently work. | @test invokelatest(getglobal, (@eval (using VersionedDep1; VersionedDep1)), :ver) == v"1.13"
# syntax.julia_version from Manifest = "1.14"
@test invokelatest(getglobal, (@eval (using VersionedDep2; VersionedDep2)), :ver) == v"1.14"
# syntax.julia_version from Manifest = "1.0" should be c... | @@ -1925,3 +1925,18 @@ module M58272_to end
@test parse(VersionNumber, read(`$(Base.julia_cmd()) --project=$(joinpath(explicit_env, "VersionedDep1")) $syntax_version_script`, String)) == v"1.13"
@test parse(VersionNumber, read(`$(Base.julia_cmd()) --project=$(joinpath(explicit_env, "VersionedDep2")) $syntax_v... | test/loading.jl | 26 | Julia | 0.214 | suggestion | 77 | 34 | 37 | false | Skip syntax check in require_stdlib. Add test. Update stdlib manifest | 60,791 | JuliaLang/julia | 35,573 | Julia | IanButterworth | IanButterworth | Review this pull request change.
Language: Julia
File: test/loading.jl
PR Title: Skip syntax check in require_stdlib. Add test. Update stdlib manifest
BEFORE CODE: @test invokelatest(getglobal, (@eval (using VersionedDep1; VersionedDep1)), :ver) == v"1.13"
# syntax.julia_version from Manifest = "1.1... | Just realized this test needs to monitor stderr. This doesn't currently work. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
typevar_names, field_names, field_types)
if inner_rewritten === doc_func
return ex
end
result = _emit_decls_with_rewritten(ctx, ex, decl_kind, func_ex, inner_rewritten)
return result !== nothing ? result : in... | ```suggestion
function rewrite_new_calls(ctx, ex, struct_name, global_struct_name,
typevar_names, field_names, field_types)
if kind(ex) == K"doc"
docs = ex[1]
ex = ex[2]
else
docs = nothing
end
if kind(ex) in KSet"global local"
stmts = SyntaxL... | # or in the case of splatnew this is enforced by the runtime.
typemax(Int)
else
numchildren(ex) - 1
end
elseif !is_leaf(ex)
minimum((_constructor_min_initialized(e) for e in children(ex)), init=typemax(Int))
else
typemax(Int)
end
end
# Let... | JuliaLowering/src/desugaring.jl | 26 | Julia | 1 | refactor | 1,369 | 51 | 51 | false | [JuliaLowering] Fix `new()` in global/local functions inside struct | 60,793 | JuliaLang/julia | 35,573 | Julia | mlechu | aviatesk | Review this pull request change.
Language: Julia
File: JuliaLowering/src/desugaring.jl
PR Title: [JuliaLowering] Fix `new()` in global/local functions inside struct
BEFORE CODE: typevar_names, field_names, field_types)
if inner_rewritten === doc_func
... | ```suggestion
function rewrite_new_calls(ctx, ex, struct_name, global_struct_name,
typevar_names, field_names, field_types)
if kind(ex) == K"doc"
docs = ex[1]
ex = ex[2]
else
docs = nothing
end
if kind(ex) in KSet"global local"
stmts = SyntaxL... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. | |
- `bail_out_apply(::AbstractInterpreter, rt, ::InferenceState)`: bail out from
`_apply_iterate` inference when return type `rt` grows up to `Any`
It also bails out from local statement/frame inference when any lattice element gets down to `Bottom`,
but `AbstractInterpreter` doesn't provide a specific interface for c... | style nit: Don't make code unnecessarily slow and complicated for no reason
```suggestion
localcache::InferenceCache
``` | `_apply_iterate` inference when return type `rt` grows up to `Any`
It also bails out from local statement/frame inference when any lattice element gets down to `Bottom`,
but `AbstractInterpreter` doesn't provide a specific interface for configuring it.
"""
function bail_out_toplevel_call end, function bail_out_call ... | @@ -509,16 +549,34 @@ typeinf_lattice(::AbstractInterpreter) = InferenceLattice(BaseInferenceLattice.i
ipo_lattice(::AbstractInterpreter) = InferenceLattice(IPOResultLattice.instance)
optimizer_lattice(::AbstractInterpreter) = SimpleInferenceLattice.instance
-struct OverlayCodeCache{Cache}
+struct OverlayCodeCache{... | Compiler/src/types.jl | 26 | Julia | 0.714 | performance | 127 | 51 | 51 | false | Use dictionary instead of linear search for the local cache. | 60,795 | JuliaLang/julia | 35,573 | Julia | vtjnash | gbaraldi | Review this pull request change.
Language: Julia
File: Compiler/src/types.jl
PR Title: Use dictionary instead of linear search for the local cache.
BEFORE CODE: - `bail_out_apply(::AbstractInterpreter, rt, ::InferenceState)`: bail out from
`_apply_iterate` inference when return type `rt` grows up to `Any`
It als... | style nit: Don't make code unnecessarily slow and complicated for no reason
```suggestion
localcache::InferenceCache
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
while (value == -1 && m->parent != m && m != jl_base_module) {
m = m->parent;
value = m->max_methods;
}
return value;
}
JL_DLLEXPORT void jl_set_istopmod(jl_module_t *self, uint8_t isprimary)
{
self->istopmod = 1;
if (isprimary) {
jl_top_module = self;
}
}
JL_DLLEXPORT ... | nit: slightly more canonical form?
```suggestion
jl_set_typetagof(g, jl_globalref_tag, 0);
``` | while (value == -1 && m->parent != m && m != jl_base_module) {
m = m->parent;
value = m->max_methods;
}
return value;
}
JL_DLLEXPORT void jl_set_istopmod(jl_module_t *self, uint8_t isprimary)
{
self->istopmod = 1;
if (isprimary) {
jl_top_module = self;
}
}
JL_DLLEXPORT ... | @@ -776,6 +776,8 @@ static jl_globalref_t *jl_new_globalref(jl_module_t *mod, jl_sym_t *name, jl_bin
{
jl_task_t *ct = jl_current_task;
jl_globalref_t *g = (jl_globalref_t*)jl_gc_alloc(ct->ptls, sizeof(jl_globalref_t), jl_globalref_type);
+ if (jl_globalref_type->smalltag)
+ jl_set_typetagof(g, jl_... | src/module.c | 26 | C | 0.714 | suggestion | 98 | 51 | 51 | false | Compiler: add smalltags for IR node types and optimize _useref_getindex | 60,831 | JuliaLang/julia | 35,573 | Julia | vtjnash | gbaraldi | Review this pull request change.
Language: C
File: src/module.c
PR Title: Compiler: add smalltags for IR node types and optimize _useref_getindex
BEFORE CODE: while (value == -1 && m->parent != m && m != jl_base_module) {
m = m->parent;
value = m->max_methods;
}
return value;
}
JL_DLLEX... | nit: slightly more canonical form?
```suggestion
jl_set_typetagof(g, jl_globalref_tag, 0);
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
1:3
```
See also [`reshape`](@ref), [`dropdims`](@ref).
"""
vec(a::AbstractArray) = reshape(a,length(a))
vec(a::AbstractVector) = a
_sub(::Tuple{}, ::Tuple{}) = ()
_sub(t::Tuple, ::Tuple{}) = t
_sub(t::Tuple, s::Tuple) = _sub(tail(t), tail(s))
"""
dropdims(A; dims)
Return an array with the same data as `A`, but... | ```suggestion
Inverse of [`insertdims`](@ref).
``` | 1:3
```
See also [`reshape`](@ref), [`dropdims`](@ref).
"""
vec(a::AbstractArray) = reshape(a,length(a))
vec(a::AbstractVector) = a
_sub(::Tuple{}, ::Tuple{}) = ()
_sub(t::Tuple, ::Tuple{}) = t
_sub(t::Tuple, s::Tuple) = _sub(tail(t), tail(s))
"""
dropdims(A; dims)
Return an array with the same data as `A`, but... | @@ -49,13 +49,16 @@ _sub(t::Tuple, s::Tuple) = _sub(tail(t), tail(s))
dropdims(A; dims)
Return an array with the same data as `A`, but with the dimensions specified by
-`dims` removed. `size(A,d)` must equal 1 for every `d` in `dims`,
-and repeated dimensions or numbers outside `1:ndims(A)` are forbidden.
+`dim... | base/abstractarraymath.jl | 26 | Julia | 0.571 | suggestion | 52 | 51 | 51 | false | Improve `insertdims` & `dropdims` docstrings | 60,833 | JuliaLang/julia | 35,573 | Julia | fingolfin | fingolfin | Review this pull request change.
Language: Julia
File: base/abstractarraymath.jl
PR Title: Improve `insertdims` & `dropdims` docstrings
BEFORE CODE: 1:3
```
See also [`reshape`](@ref), [`dropdims`](@ref).
"""
vec(a::AbstractArray) = reshape(a,length(a))
vec(a::AbstractVector) = a
_sub(::Tuple{}, ::Tuple{}) = ()
_... | ```suggestion
Inverse of [`insertdims`](@ref).
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# This file is a part of Julia. License is MIT: https://julialang.org/license
module TestDownload
using Test
mktempdir() do temp_dir
try
download("https://httpbin.julialang.org")
global url = "https://httpbin.julialang.org"
catch
global url = "https://httpbin.org"
end
# Downlo... | ```suggestion
url = try
download("https://httpbin.julialang.org")
"https://httpbin.julialang.org"
catch
"https://httpbin.org"
end
```
Is `global` necessary? | # This file is a part of Julia. License is MIT: https://julialang.org/license
module TestDownload
using Test
mktempdir() do temp_dir
url = try
download("https://httpbingo.julialang.org")
"https://httpbingo.julialang.org"
catch ex
bt = catch_backtrace()
@info "Looks like there ... | @@ -5,22 +5,28 @@ module TestDownload
using Test
mktempdir() do temp_dir
+ try
+ download("https://httpbin.julialang.org")
+ global url = "https://httpbin.julialang.org"
+ catch
+ global url = "https://httpbin.org"
+ end | test/download_exec.jl | 13 | Julia | 0.714 | question | 192 | 38 | 38 | false | Test suite: switch to `httpbingo.julialang`, fall back to `httpbin.julialang` on failure | 60,834 | JuliaLang/julia | 35,573 | Julia | giordano | rokke-git | Review this pull request change.
Language: Julia
File: test/download_exec.jl
PR Title: Test suite: switch to `httpbingo.julialang`, fall back to `httpbin.julialang` on failure
BEFORE CODE: # This file is a part of Julia. License is MIT: https://julialang.org/license
module TestDownload
using Test
mktempdir() do ... | ```suggestion
url = try
download("https://httpbin.julialang.org")
"https://httpbin.julialang.org"
catch
"https://httpbin.org"
end
```
Is `global` necessary? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# This file is a part of Julia. License is MIT: https://julialang.org/license
module TestDownload
using Test
mktempdir() do temp_dir
url = try
download("https://httpbin.julialang.org")
"https://httpbin.julialang.org"
catch
"https://httpbin.org"
end
# Download a file
file =... | ```suggestion
"https://httpbingo.org"
``` | # This file is a part of Julia. License is MIT: https://julialang.org/license
module TestDownload
using Test
mktempdir() do temp_dir
url = try
download("https://httpbingo.julialang.org")
"https://httpbingo.julialang.org"
catch ex
bt = catch_backtrace()
@info "Looks like there ... | @@ -5,22 +5,28 @@ module TestDownload
using Test
mktempdir() do temp_dir
+ url = try
+ download("https://httpbin.julialang.org")
+ "https://httpbin.julialang.org"
+ catch
+ "https://httpbin.org" | test/download_exec.jl | 12 | Julia | 0.429 | suggestion | 49 | 37 | 37 | false | Test suite: switch to `httpbingo.julialang`, fall back to `httpbin.julialang` on failure | 60,834 | JuliaLang/julia | 35,573 | Julia | inkydragon | rokke-git | Review this pull request change.
Language: Julia
File: test/download_exec.jl
PR Title: Test suite: switch to `httpbingo.julialang`, fall back to `httpbin.julialang` on failure
BEFORE CODE: # This file is a part of Julia. License is MIT: https://julialang.org/license
module TestDownload
using Test
mktempdir() do ... | ```suggestion
"https://httpbingo.org"
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# This file is a part of Julia. License is MIT: https://julialang.org/license
module TestDownload
using Test
mktempdir() do temp_dir
url = try
download("https://httpbin.julialang.org")
"https://httpbin.julialang.org"
catch
"https://httpbingo.julialang.org/"
end
# Download a fi... | ```suggestion
catch ex
bt = catch_backtrace()
@info "Looks like there is a problem with a JuliaLang mirror of httpbin(go)" exception=(ex,bt)
"https://httpbingo.julialang.org/"
``` | # This file is a part of Julia. License is MIT: https://julialang.org/license
module TestDownload
using Test
mktempdir() do temp_dir
url = try
download("https://httpbingo.julialang.org")
"https://httpbingo.julialang.org"
catch ex
bt = catch_backtrace()
@info "Looks like there ... | @@ -5,22 +5,28 @@ module TestDownload
using Test
mktempdir() do temp_dir
+ url = try
+ download("https://httpbin.julialang.org")
+ "https://httpbin.julialang.org"
+ catch
+ "https://httpbingo.julialang.org/" | test/download_exec.jl | 12 | Julia | 0.929 | suggestion | 212 | 37 | 37 | false | Test suite: switch to `httpbingo.julialang`, fall back to `httpbin.julialang` on failure | 60,834 | JuliaLang/julia | 35,573 | Julia | DilumAluthge | rokke-git | Review this pull request change.
Language: Julia
File: test/download_exec.jl
PR Title: Test suite: switch to `httpbingo.julialang`, fall back to `httpbin.julialang` on failure
BEFORE CODE: # This file is a part of Julia. License is MIT: https://julialang.org/license
module TestDownload
using Test
mktempdir() do ... | ```suggestion
catch ex
bt = catch_backtrace()
@info "Looks like there is a problem with a JuliaLang mirror of httpbin(go)" exception=(ex,bt)
"https://httpbingo.julialang.org/"
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# This file is a part of Julia. License is MIT: https://julialang.org/license
module TestDownload
using Test
mktempdir() do temp_dir
url = try
download("https://httpbin.julialang.org")
"https://httpbin.julialang.org"
catch ex
bt = catch_backtrace()
@info "Looks like there is a... | Let's start with `httpbingo`, and then fall back to `httpbin`.
```suggestion
url = try
download("https://httpbingo.julialang.org")
"https://httpbingo.julialang.org"
catch ex
bt = catch_backtrace()
@info "Looks like there is a problem with a JuliaLang mirror of httpbingo"... | # This file is a part of Julia. License is MIT: https://julialang.org/license
module TestDownload
using Test
mktempdir() do temp_dir
url = try
download("https://httpbingo.julialang.org")
"https://httpbingo.julialang.org"
catch ex
bt = catch_backtrace()
@info "Looks like there ... | @@ -5,22 +5,31 @@ module TestDownload
using Test
mktempdir() do temp_dir
+ url = try
+ download("https://httpbin.julialang.org")
+ "https://httpbin.julialang.org"
+ catch ex
+ bt = catch_backtrace()
+ @info "Looks like there is a problem with a JuliaLang mirror of httpbin."
+ ... | test/download_exec.jl | 16 | Julia | 0.929 | suggestion | 426 | 41 | 41 | false | Test suite: switch to `httpbingo.julialang`, fall back to `httpbin.julialang` on failure | 60,834 | JuliaLang/julia | 35,573 | Julia | DilumAluthge | rokke-git | Review this pull request change.
Language: Julia
File: test/download_exec.jl
PR Title: Test suite: switch to `httpbingo.julialang`, fall back to `httpbin.julialang` on failure
BEFORE CODE: # This file is a part of Julia. License is MIT: https://julialang.org/license
module TestDownload
using Test
mktempdir() do ... | Let's start with `httpbingo`, and then fall back to `httpbin`.
```suggestion
url = try
download("https://httpbingo.julialang.org")
"https://httpbingo.julialang.org"
catch ex
bt = catch_backtrace()
@info "Looks like there is a problem with a JuliaLang mirror of httpbingo"... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
exit_on_error::Bool=false,
revise::Bool=false,
propagate_project::Bool=false,
seed::Union{BitInteger,Nothing}=nothing,
julia_args::Cmd=``)
if isa(tests,AbstractString)
tests = split(tests)
end
exit_on_error && ... | ```suggestion
vinfo = read(setenv(`$(julia_cmd()) -e 'let InteractiveUtils = Base.require_stdlib(Base.PkgId(Base.UUID(0xb77e0a4c_d291_57a0_90e8_8db25a27a240), "InteractiveUtils")); @invokelatest(InteractiveUtils.versioninfo()); end'`, ENV2), String)
```
| exit_on_error::Bool=false,
revise::Bool=false,
propagate_project::Bool=false,
seed::Union{BitInteger,Nothing}=nothing,
julia_args::Cmd=``)
if isa(tests,AbstractString)
tests = split(tests)
end
exit_on_error && ... | @@ -719,12 +719,16 @@ function runtests(tests = ["all"]; ncores::Int = ceil(Int, Sys.EFFECTIVE_CPU_THR
Base.DATAROOTDIR, "julia", "test", "runtests.jl")) $tests`, ENV2))
nothing
catch
- buf = PipeBuffer()
- let InteractiveUtils = Base.require_stdlib(PkgId(UUID(0xb77e0a4c_d291_57... | base/util.jl | 26 | Julia | 0.786 | suggestion | 262 | 46 | 46 | false | Clean up Base.runtests failure print | 60,837 | JuliaLang/julia | 35,573 | Julia | IanButterworth | IanButterworth | Review this pull request change.
Language: Julia
File: base/util.jl
PR Title: Clean up Base.runtests failure print
BEFORE CODE: exit_on_error::Bool=false,
revise::Bool=false,
propagate_project::Bool=false,
seed::Union{BitInteger,Nothing}=nothin... | ```suggestion
vinfo = read(setenv(`$(julia_cmd()) -e 'let InteractiveUtils = Base.require_stdlib(Base.PkgId(Base.UUID(0xb77e0a4c_d291_57a0_90e8_8db25a27a240), "InteractiveUtils")); @invokelatest(InteractiveUtils.versioninfo()); end'`, ENV2), String)
```
| You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
function provenance(ex::SyntaxTree)
s = ex.source
if s isa NodeId
return (SyntaxTree(ex._graph, s),)
elseif s isa Tuple
return SyntaxTree.((ex._graph,), s)
else
return (s,)
end
end
function _sourceref(sources, id)
i = 1
while true
i += 1
s = sources[i... | Should we try using `Dict{NodeId}` in more places? Right now I believe we're using `Dict{Any, Any}` (inside the `Dict{Symbol, X}`) everywhere. Not sure it would change much. | function provenance(ex::SyntaxTree)
s = ex.source
if s isa NodeId
return (SyntaxTree(ex._graph, s),)
elseif s isa Tuple
return SyntaxTree.((ex._graph,), s)
else
return (s,)
end
end
function _sourceref(sources, id)
i = 1
while true
i += 1
s = sources[i... | @@ -371,8 +370,8 @@ function _sourceref(sources, id)
end
function sourceref(ex::SyntaxTree)
- sources = ex._graph.source
- id::NodeId = ex._id
+ sources = ex._graph.source::Dict{NodeId} | JuliaSyntax/src/porcelain/syntax_graph.jl | 26 | Julia | 0.571 | question | 174 | 51 | 51 | false | [JuliaSyntax] improve type stability of `SyntaxTree` query APIs | 60,851 | JuliaLang/julia | 35,573 | Julia | mlechu | aviatesk | Review this pull request change.
Language: Julia
File: JuliaSyntax/src/porcelain/syntax_graph.jl
PR Title: [JuliaSyntax] improve type stability of `SyntaxTree` query APIs
BEFORE CODE: function provenance(ex::SyntaxTree)
s = ex.source
if s isa NodeId
return (SyntaxTree(ex._graph, s),)
elseif s is... | Should we try using `Dict{NodeId}` in more places? Right now I believe we're using `Dict{Any, Any}` (inside the `Dict{Symbol, X}`) everywhere. Not sure it would change much. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
"|(&\nfunction" => "(call | (& (function (error (error)) (block (error)) (error-t))) (error-t))"
"@(" => "(macrocall (macro_name (parens (error-t))))"
"x = @(" => "(= x (macrocall (macro_name (parens (error-t)))))"
"function(where" => "(function (tuple-p where (error-t)) (block (error)) (error-t))"
... | I think we should disallow this one in 1.14 syntax - needs to work in 1.13 syntax of course, but i think it's too confusing otherwise and we may want it for more complicated labels in the future. | ((v=v"1.14",), "break f()") => "(break f (error-t) (tuple-p))"
((v=v"1.14",), "continue _") => "(continue _)"
((v=v"1.14",), "continue label") => "(continue label)"
((v=v"1.14",), "break +") => "(break (error-t +))"
((v=v"1.14",), "a[break label end]") => "(ref a (break label end))"
... | @@ -1127,6 +1133,13 @@ parsestmt_test_specs = [
# unit tests there.
"''" => "(char (error))"
+ # break / continue with trailing tokens are legal in some cases
+ "a ? break : c" => "(? a (break) c)"
+ "begin break end" => "(block (break))"
+ "a ? continue : c" => "(? a (continue) c)"
+ "... | JuliaSyntax/test/parser.jl | 26 | Julia | 0.429 | suggestion | 195 | 51 | 51 | false | Fixes for `break label` and `continue label` parsing | 60,910 | JuliaLang/julia | 35,573 | Julia | Keno | c42f | Review this pull request change.
Language: Julia
File: JuliaSyntax/test/parser.jl
PR Title: Fixes for `break label` and `continue label` parsing
BEFORE CODE: "|(&\nfunction" => "(call | (& (function (error (error)) (block (error)) (error-t))) (error-t))"
"@(" => "(macrocall (macro_name (parens (error-t))))"... | I think we should disallow this one in 1.14 syntax - needs to work in 1.13 syntax of course, but i think it's too confusing otherwise and we may want it for more complicated labels in the future. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@test !global_assigned
end
@testset "soft scope isn't top level" begin
ex = quote
begin
for i in 1:1; global soft_assigned_explicit_global = 1; end
for i in 1:1; soft_assigned_explicit_global = 2; end
end
end
expr_eval(test... | Might be nice to encode these skip / error cases into `results`, but pretty orthogonal to this PR | (true, decls_neutral, wrap_none, wrap_neutral) => (true,true),
(true, decls_none, wrap_hard, wrap_func ) => (true,false),
(true, decls_none, wrap_hard, wrap_hard ) => (true,false),
(true, decls_none, wrap_hard, wrap_neutral) => (true,false),
(true, d... | @@ -147,85 +147,136 @@ end
# Switch to Core.eval for sanity-checking
expr_eval(mod, ex) = JuliaLowering.eval(mod, ex)
-enable_softscope(e) = Expr(:block, Expr(:softscope, true), e)
-inner_neutral(e) = :(for _ in 1:1; $e; end)
-inner_func(e) = :((function (); $e end)(#=called=#))
-inner_hard(e) = :(let; $e end)
-
-o... | JuliaLowering/test/scopes.jl | 26 | Julia | 0.5 | suggestion | 97 | 51 | 51 | false | [JuliaLowering] Fix soft scope semantics leaking through hard scope | 60,917 | JuliaLang/julia | 35,573 | Julia | topolarity | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/test/scopes.jl
PR Title: [JuliaLowering] Fix soft scope semantics leaking through hard scope
BEFORE CODE: @test !global_assigned
end
@testset "soft scope isn't top level" begin
ex = quote
begin
... | Might be nice to encode these skip / error cases into `results`, but pretty orthogonal to this PR | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
(a->prefer_specsig == b->prefer_specsig) &&
(a->gnu_pubnames == b->gnu_pubnames) &&
(a->debug_info_kind == b->debug_info_kind) &&
(a->safepoint_on_entry == b->safepoint_on_entry) &&
(a->gcstack_arg == b->gcstack_arg) &&
(a->use_jlplt == b->use_jlplt) &&
... | fixme: this probably shouldn't be hard coded | (a->prefer_specsig == b->prefer_specsig) &&
(a->gnu_pubnames == b->gnu_pubnames) &&
(a->debug_info_kind == b->debug_info_kind) &&
(a->safepoint_on_entry == b->safepoint_on_entry) &&
(a->gcstack_arg == b->gcstack_arg) &&
(a->use_jlplt == b->use_jlplt) &&
... | @@ -4681,18 +4725,37 @@ static int compare_cgparams(const jl_cgparams_t *a, const jl_cgparams_t *b)
}
#endif
-static auto *emit_genericmemory_unchecked(jl_codectx_t &ctx, Value *cg_nbytes, Value *cg_typ)
+// Emit allocation for variable-length GenericMemory
+// If zeroinit_nbytes is non-null, adds a zeroinit_indire... | src/cgutils.cpp | 26 | C++ | 0.143 | suggestion | 44 | 51 | 51 | false | codegen: Move GC pointer zeroing to late-gc-lowering | 60,924 | JuliaLang/julia | 35,573 | Julia | gbaraldi | gbaraldi | Review this pull request change.
Language: C++
File: src/cgutils.cpp
PR Title: codegen: Move GC pointer zeroing to late-gc-lowering
BEFORE CODE: (a->prefer_specsig == b->prefer_specsig) &&
(a->gnu_pubnames == b->gnu_pubnames) &&
(a->debug_info_kind == b->debug_info_kind) &&
... | fixme: this probably shouldn't be hard coded | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# TODO: metas
is_literal(k) || is_identifier_like(ex) || k == K"Symbol" ||
k == K"inert" || k == K"inert_syntaxtree" || k == K"top" ||
k == K"core" || k == K"Value"
# flisp also includes `a.b` with simple `a`, but this seems like a bug
# because this calls the user-defined getproperty?
e... | Looks reasonable. Hopefully it's not a problem to have a *lot* of scope layers... they are lighter than `BindingInfo` though so I guess it's fine!
It would be nice to ensure these never end up global "by mistake" in the same way that macro expanded identifiers skip accidentally becoming global by the use of the is_mac... | # TODO: metas
is_literal(k) || is_identifier_like(ex) || k == K"Symbol" ||
k == K"inert" || k == K"inert_syntaxtree" || k == K"top" ||
k == K"core" || k == K"Value"
# flisp also includes `a.b` with simple `a`, but this seems like a bug
# because this calls the user-defined getproperty?
e... | @@ -93,6 +93,13 @@ function check_no_assignment(exs, msg="misplaced assignment statement in `[ ...
end
end
+# Generating a new_local_binding or ssaval should only be done if we can
+# guarantee there's some scope it's declared in, and that it's not declared or
+# used outside of that scope (binding capture is O... | JuliaLowering/src/desugaring.jl | 26 | Julia | 0.643 | suggestion | 562 | 51 | 51 | false | [JuliaLowering] Make locals introduced in `K"method_defs"` top-level | 60,942 | JuliaLang/julia | 35,573 | Julia | c42f | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/desugaring.jl
PR Title: [JuliaLowering] Make locals introduced in `K"method_defs"` top-level
BEFORE CODE: # TODO: metas
is_literal(k) || is_identifier_like(ex) || k == K"Symbol" ||
k == K"inert" || k == K"inert_syntaxtree" || k... | Looks reasonable. Hopefully it's not a problem to have a *lot* of scope layers... they are lighter than `BindingInfo` though so I guess it's fine!
It would be nice to ensure these never end up global "by mistake" in the same way that macro expanded identifiers skip accidentally becoming global by the use of the is_mac... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
kwcall_arg_names, kwcall_arg_types, first_default, arg_defaults)
end
positional_forwarding_args = if has_slurp
a = copy(arg_names)
a[end] = @ast ctx a[end] [K"..." a[end]]
a
else
arg_names
end
#--------------------------------------... | Why the mutating variant?
In general I feel we should avoid expression mutation in lowering if possible outside of utilities in ast.jl. Yes, even in trivial/local cases because using mutation just makes everything harder to reason about.
I see there's quite a lot now... Can we do less of it, not more? If it's just a... | optional_positional_defs!(ctx, kwcall_method_defs, srcref, callex_srcref,
kwcall_mtable, typevar_names, typevar_stmts,
kwcall_arg_names, kwcall_arg_types, first_default, arg_defaults)
end
positional_forwarding_args = if has_slurp
... | @@ -2758,7 +2757,8 @@ function keyword_function_defs(ctx, srcref, callex_srcref, name_str, typevar_nam
for n in kw_names
# If not using slots for the keyword argument values, still declare
# them for reflection purposes.
- push!(kw_val_stmts, @ast ctx n [K"local"(meta=Compi... | JuliaLowering/src/desugaring.jl | 26 | Julia | 0.643 | refactor | 446 | 51 | 51 | false | [JuliaLowering] Make locals introduced in `K"method_defs"` top-level | 60,942 | JuliaLang/julia | 35,573 | Julia | c42f | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/desugaring.jl
PR Title: [JuliaLowering] Make locals introduced in `K"method_defs"` top-level
BEFORE CODE: kwcall_arg_names, kwcall_arg_types, first_default, arg_defaults)
end
positional_forwarding_arg... | Why the mutating variant?
In general I feel we should avoid expression mutation in lowering if possible outside of utilities in ast.jl. Yes, even in trivial/local cases because using mutation just makes everything harder to reason about.
I see there's quite a lot now... Can we do less of it, not more? If it's just a... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
end
top_scope(ctx) = ctx.scopes[1]
is_top_scope(scope::ScopeInfo) = scope.parent_id === 0
enclosing_lambda(ctx, scope::ScopeInfo) = ctx.scopes[scope.lambda_id]
parent(ctx, scope::ScopeInfo) = is_top_scope(scope) ? nothing :
ctx.scopes[scope.parent_id]
_var_str(v) = v === :local ? "local variable" :
v === :glo... | I assume this would be an internal error in lowering? |
top_scope(ctx) = ctx.scopes[1]
is_top_scope(scope::ScopeInfo) = scope.parent_id === 0
enclosing_lambda(ctx, scope::ScopeInfo) = ctx.scopes[scope.lambda_id]
parent(ctx, scope::ScopeInfo) = is_top_scope(scope) ? nothing :
ctx.scopes[scope.parent_id]
_var_str(v) = v === :local ? "local variable" :
v === :global ... | @@ -113,8 +126,12 @@ _var_str(v) = v === :local ? "local variable" :
function maybe_declare_in_scope!(ctx, scope::ScopeInfo, ex, new_k::Symbol)
if kind(ex) === K"BindingId"
bid = ex.var_id
- @assert get_binding(ctx, bid).kind === new_k
- record_lambda_var!(ctx, scope, get_binding(ctx, bid),... | JuliaLowering/src/scope_analysis.jl | 26 | Julia | 0.357 | question | 53 | 51 | 51 | false | [JuliaLowering] Make locals introduced in `K"method_defs"` top-level | 60,942 | JuliaLang/julia | 35,573 | Julia | c42f | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/scope_analysis.jl
PR Title: [JuliaLowering] Make locals introduced in `K"method_defs"` top-level
BEFORE CODE: end
top_scope(ctx) = ctx.scopes[1]
is_top_scope(scope::ScopeInfo) = scope.parent_id === 0
enclosing_lambda(ctx, scope::ScopeInfo) = ... | I assume this would be an internal error in lowering? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
end
elseif old_k === new_k
(new_k === :global || new_k === :local) && return bid
throw(LoweringError(ex, "function $(_var_str(new_k)) name not unique"))
# See note in test/scopes.jl: "globals may overlap args or sparams"
# elseif new_k === :global && old_k in (:argument, :static_para... | Looks reasonable. It could also be ok or maybe good to have an extra flag on `ScopeLayer` for `is_internal`. A bit like a "light" version of `BindingInfo`. | elseif old_k === new_k
(new_k === :global || new_k === :local) && return bid
throw(LoweringError(ex, "function $(_var_str(new_k)) name not unique"))
# See note in test/scopes.jl: "globals may overlap args or sparams"
# elseif new_k === :global && old_k in (:argument, :static_parameter)
#... | @@ -152,29 +169,38 @@ function declare_in_scope!(ctx, scope::ScopeInfo, ex, bk::Symbol; kws...)
declaration_scope = scope
mod = nothing
end
- b = _new_binding(ctx, ex, nk.name, bk; mod, kws...)
+ b = _new_binding(ctx, ex, nk.name, bk;
+ mod, is_internal=getmeta(ex, :is_i... | JuliaLowering/src/scope_analysis.jl | 26 | Julia | 0.571 | suggestion | 155 | 51 | 51 | false | [JuliaLowering] Make locals introduced in `K"method_defs"` top-level | 60,942 | JuliaLang/julia | 35,573 | Julia | c42f | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/scope_analysis.jl
PR Title: [JuliaLowering] Make locals introduced in `K"method_defs"` top-level
BEFORE CODE: end
elseif old_k === new_k
(new_k === :global || new_k === :local) && return bid
throw(LoweringError(ex, ... | Looks reasonable. It could also be ok or maybe good to have an extra flag on `ScopeLayer` for `is_internal`. A bit like a "light" version of `BindingInfo`. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
lam.locals_capt[b.id] = false
b.lambda_id = lam.id
nothing
end
function ensure_captured!(ctx, scope::ScopeInfo, b)
if b.kind === :global || b.is_ssa
return
end
lam = enclosing_lambda(ctx, scope)
if !haskey(lam.locals_capt, b.id)
b.is_captured = true
lam.locals_capt[b... | Sure, could easily have a lazily filled cache on the scope if we find it helps performance. | b.lambda_id = lam.id
nothing
end
function ensure_captured!(ctx, scope::ScopeInfo, b)
if b.kind === :global || b.is_ssa
return
end
lam = enclosing_lambda(ctx, scope)
if !haskey(lam.locals_capt, b.id)
b.is_captured = true
lam.locals_capt[b.id] = true
s2 = parent(ct... | @@ -183,6 +209,7 @@ function needs_resolution(ex)
end
function resolve_name(ctx, ex; exclude_toplevel_globals=false)
+ # TODO: probably want to cache these lookups | JuliaLowering/src/scope_analysis.jl | 26 | Julia | 0.429 | performance | 91 | 51 | 51 | false | [JuliaLowering] Make locals introduced in `K"method_defs"` top-level | 60,942 | JuliaLang/julia | 35,573 | Julia | c42f | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/scope_analysis.jl
PR Title: [JuliaLowering] Make locals introduced in `K"method_defs"` top-level
BEFORE CODE: lam.locals_capt[b.id] = false
b.lambda_id = lam.id
nothing
end
function ensure_captured!(ctx, scope::ScopeInfo, b)
i... | Sure, could easily have a lazily filled cache on the scope if we find it helps performance. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
6 (isdefined slot₁/y)
7 (gotoifnot %₆ label₁₀)
8 slot₁/y
9 (call top.setindex! %₅ %₈ :y)
10 (return %₅)
########################################
# @locals with function args (TODO: static parameters)
function f(z)
@locals
end
#---------------------
1 (method TestMod.f)
2 latestworld
3 TestMod.f
4 ... | `@locals` should not (or currently does not, in flisp) include internal names so I think this is a bug. | 6 (isdefined slot₁/y)
7 (gotoifnot %₆ label₁₀)
8 slot₁/y
9 (call top.setindex! %₅ %₈ :y)
10 (return %₅)
########################################
# @locals with function args (TODO: static parameters)
function f(z)
@locals
end
#---------------------
1 (method TestMod.f)
2 latestworld
3 TestMod.f
4 ... | @@ -199,12 +199,14 @@ end
7 SourceLocation::1:10
8 (call core.svec %₅ %₆ %₇)
9 --- method core.nothing %₈
- slots: [slot₁/#self#(!read) slot₂/z]
+ slots: [slot₁/#self# slot₂/z]
1 (call core.apply_type top.Dict core.Symbol core.Any)
2 (call %₁)
3 (gotoifnot true label₅)
- 4 (call... | JuliaLowering/test/scopes_ir.jl | 26 | Julia | 0.714 | bug | 103 | 51 | 51 | false | [JuliaLowering] Make locals introduced in `K"method_defs"` top-level | 60,942 | JuliaLang/julia | 35,573 | Julia | c42f | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/test/scopes_ir.jl
PR Title: [JuliaLowering] Make locals introduced in `K"method_defs"` top-level
BEFORE CODE: 6 (isdefined slot₁/y)
7 (gotoifnot %₆ label₁₀)
8 slot₁/y
9 (call top.setindex! %₅ %₈ :y)
10 (return %₅)
#######################... | `@locals` should not (or currently does not, in flisp) include internal names so I think this is a bug. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
function inner(x::T)::Tuple{T, Vector{T}} where T
(x,T[x])
end
inner(a)
end
f_inner_rt_sp(1), f_inner_rt_sp("foo")
end
""") == ((1, [1]), ("foo", ["foo"]))
@test JuliaLowering.include_string(test_mod, """
begin
function f_many_closure_sp()
function (); function (... | I don't expect this test to work: the `T` is not a typevar which is in scope for the inner closure, it's just a part of the `UnionAll` definition?
Did you mean to use `f_many_closure_sp_capt(x::T) where T` ? | function inner(x::T)::Tuple{T, Vector{T}} where T
(x,T[x])
end
inner(a)
end
f_inner_rt_sp(1), f_inner_rt_sp("foo")
end
""") == ((1, [1]), ("foo", ["foo"]))
@test JuliaLowering.include_string(test_mod, """
begin
function f_many_closure_sp()
function (); function (... | @@ -472,6 +472,81 @@ end
@test_throws UndefVarError test_mod.f_undef_static_param(nothing)()
@test test_mod.f_undef_static_param(42)() == Int
+@test JuliaLowering.include_string(test_mod, """
+begin
+ function f_inner_sp(x::T) where T
+ function inner(y::U) where U
+ (T, U)
+ end
+ ... | JuliaLowering/test/closures.jl | 26 | Julia | 0.786 | question | 208 | 51 | 51 | false | [JuliaLowering] Make locals introduced in `K"method_defs"` top-level | 60,942 | JuliaLang/julia | 35,573 | Julia | c42f | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/test/closures.jl
PR Title: [JuliaLowering] Make locals introduced in `K"method_defs"` top-level
BEFORE CODE: function inner(x::T)::Tuple{T, Vector{T}} where T
(x,T[x])
end
inner(a)
end
f_inner_rt_sp(1), ... | I don't expect this test to work: the `T` is not a typevar which is in scope for the inner closure, it's just a part of the `UnionAll` definition?
Did you mean to use `f_many_closure_sp_capt(x::T) where T` ? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@test JuliaLowering.include_string(test_mod, """
begin
function f_many_closure_sp()
function (); function (); function (x::T) where T; (x, T) end; end; end
end
f_many_closure_sp()()()(1)
end
""") == (1, Int)
@test_broken JuliaLowering.include_string(test_mod, """
begin
function f_many_closure_s... | This one seems to be a bug, yes.
I thought I tried to make this work. But maybe I'm just remembering reading the flisp code which goes to extra lengths to get selected cases working when inner method declarations depend on static parameters from the outer function.
If I remember flisp makes this work by making the ty... | @test JuliaLowering.include_string(test_mod, """
begin
function f_many_closure_sp()
function (); function (); function (x::T) where T; (x, T) end; end; end
end
f_many_closure_sp()()()(1)
end
""") == (1, Int)
@test JuliaLowering.include_string(test_mod, """
begin
function f_many_closure_sp_capt(... | @@ -472,6 +472,81 @@ end
@test_throws UndefVarError test_mod.f_undef_static_param(nothing)()
@test test_mod.f_undef_static_param(42)() == Int
+@test JuliaLowering.include_string(test_mod, """
+begin
+ function f_inner_sp(x::T) where T
+ function inner(y::U) where U
+ (T, U)
+ end
+ ... | JuliaLowering/test/closures.jl | 26 | Julia | 0.714 | bug | 351 | 51 | 51 | false | [JuliaLowering] Make locals introduced in `K"method_defs"` top-level | 60,942 | JuliaLang/julia | 35,573 | Julia | c42f | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/test/closures.jl
PR Title: [JuliaLowering] Make locals introduced in `K"method_defs"` top-level
BEFORE CODE: @test JuliaLowering.include_string(test_mod, """
begin
function f_many_closure_sp()
function (); function (); function (x::T) ... | This one seems to be a bug, yes.
I thought I tried to make this work. But maybe I'm just remembering reading the flisp code which goes to extra lengths to get selected cases working when inner method declarations depend on static parameters from the outer function.
If I remember flisp makes this work by making the ty... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
function f_many_closure_sp()
function (); function (); function (x::T) where T; (x, T) end; end; end
end
f_many_closure_sp()()()(1)
end
""") == (1, Int)
@test_broken JuliaLowering.include_string(test_mod, """
begin
function f_many_closure_sp_capt(x::T where T)
function (); function (); ... | ```suggestion
# Inner method typevar `U` depending on a static parameter `T` so hoisting the method def
# for `inner` out to top level would require detecting this and making `inner` parametric on `T`.
# Note this doesn't work in flisp either
``` | function f_many_closure_sp()
function (); function (); function (x::T) where T; (x, T) end; end; end
end
f_many_closure_sp()()()(1)
end
""") == (1, Int)
@test JuliaLowering.include_string(test_mod, """
begin
function f_many_closure_sp_capt(x::T) where T
function (); function (); functio... | @@ -472,6 +472,81 @@ end
@test_throws UndefVarError test_mod.f_undef_static_param(nothing)()
@test test_mod.f_undef_static_param(42)() == Int
+@test JuliaLowering.include_string(test_mod, """
+begin
+ function f_inner_sp(x::T) where T
+ function inner(y::U) where U
+ (T, U)
+ end
+ ... | JuliaLowering/test/closures.jl | 26 | Julia | 0.929 | suggestion | 246 | 51 | 51 | false | [JuliaLowering] Make locals introduced in `K"method_defs"` top-level | 60,942 | JuliaLang/julia | 35,573 | Julia | c42f | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/test/closures.jl
PR Title: [JuliaLowering] Make locals introduced in `K"method_defs"` top-level
BEFORE CODE: function f_many_closure_sp()
function (); function (); function (x::T) where T; (x, T) end; end; end
end
f_many_closur... | ```suggestion
# Inner method typevar `U` depending on a static parameter `T` so hoisting the method def
# for `inner` out to top level would require detecting this and making `inner` parametric on `T`.
# Note this doesn't work in flisp either
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
########################################
# Error: ccall with too many arguments
ccall(:strlen, Csize_t, (Cstring,), "asdfg", "blah")
#---------------------
LoweringError:
ccall(:strlen, Csize_t, (Cstring,), "asdfg", "blah")
└──────────────────────────────────────────────────┘ ── More arguments than types in ccall
####... | ```suggestion
# cglobal special support for (sym, lib) tuple
# unlike flisp we outline the tuple and allow constant propgation to put it back
# before codegen generates code for `cglobal`
``` | ########################################
# Error: ccall with too many arguments
ccall(:strlen, Csize_t, (Cstring,), "asdfg", "blah")
#---------------------
LoweringError:
ccall(:strlen, Csize_t, (Cstring,), "asdfg", "blah")
└──────────────────────────────────────────────────┘ ── More arguments than types in ccall
####... | @@ -502,9 +502,11 @@ ccall(:foo, Csize_t, (Cstring..., Cstring...), "asdfg", "blah")
# cglobal special support for (sym, lib) tuple | JuliaLowering/test/function_calls_ir.jl | 26 | Julia | 0.714 | suggestion | 191 | 51 | 51 | false | [JuliaLowering] Make locals introduced in `K"method_defs"` top-level | 60,942 | JuliaLang/julia | 35,573 | Julia | c42f | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/test/function_calls_ir.jl
PR Title: [JuliaLowering] Make locals introduced in `K"method_defs"` top-level
BEFORE CODE: ########################################
# Error: ccall with too many arguments
ccall(:strlen, Csize_t, (Cstring,), "asdfg", "bla... | ```suggestion
# cglobal special support for (sym, lib) tuple
# unlike flisp we outline the tuple and allow constant propgation to put it back
# before codegen generates code for `cglobal`
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
{
Value *bp = julia_binding_gv(ctx, bnd);
Value *v = ctx.builder.CreateCall(prepare_call(jlgetbindingvalue_func), { bp });
undef_var_error_ifnot(ctx, ctx.builder.CreateIsNotNull(v), name, (jl_value_t*)mod);
return mark_julia_type(ctx, v, true, jl_any_type);
}
static jl_cgval_t emit_globalref(jl_codectx... | ```suggestion
if (jl_generating_output())
jl_temporary_root(ctx, constval);
``` | {
Value *bp = julia_binding_gv(ctx, bnd);
Value *v = ctx.builder.CreateCall(prepare_call(jlgetbindingvalue_func), { bp });
undef_var_error_ifnot(ctx, ctx.builder.CreateIsNotNull(v), name, (jl_value_t*)mod);
return mark_julia_type(ctx, v, true, jl_any_type);
}
static jl_cgval_t emit_globalref(jl_codectx... | @@ -3505,6 +3505,8 @@ static jl_cgval_t emit_globalref(jl_codectx_t &ctx, jl_module_t *mod, jl_sym_t *
undef_var_error_ifnot(ctx, ConstantInt::get(getInt1Ty(ctx.builder.getContext()), 0), name, (jl_value_t*)mod);
return jl_cgval_t();
}
+ // root is required to allow bindings to... | src/codegen.cpp | 26 | C++ | 0.571 | suggestion | 99 | 51 | 51 | false | codegen: Mark resolved `const` `GlobalRef` as method root | 60,945 | JuliaLang/julia | 35,573 | Julia | vtjnash | topolarity | Review this pull request change.
Language: C++
File: src/codegen.cpp
PR Title: codegen: Mark resolved `const` `GlobalRef` as method root
BEFORE CODE: {
Value *bp = julia_binding_gv(ctx, bnd);
Value *v = ctx.builder.CreateCall(prepare_call(jlgetbindingvalue_func), { bp });
undef_var_error_ifnot(ctx, ctx.... | ```suggestion
if (jl_generating_output())
jl_temporary_root(ctx, constval);
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@test setindex!(fill(0.0), fill(1.0), CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, :) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0, CI0) == fill(1.0)
... | fix the test
```suggestion
@test Base.infer_return_type(checkbounds, Tuple{typ, Any}) <: Nothing
``` | @test setindex!(fill(0.0), fill(1.0), CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, :) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0, CI0) == fill(1.0)
... | @@ -3256,6 +3256,12 @@ end
@test setindex!(zeros(2,2), fill(1.0), CI0, 1, 1) == [1.0 0.0; 0.0 0.0]
end
+@testset "conditionally-throwing version of `checkbounds` should return `nothing` if it returns" begin
+ for typ in (AbstractString, DenseArray, Base.AbstractBroadcasted)
+ @test Base.infer_return_t... | test/arrayops.jl | 26 | Julia | 0.857 | suggestion | 109 | 51 | 51 | false | make `checkbounds(collection, indices...)` return `nothing` when returning | 60,956 | JuliaLang/julia | 35,573 | Julia | nsajko | nsajko | Review this pull request change.
Language: Julia
File: test/arrayops.jl
PR Title: make `checkbounds(collection, indices...)` return `nothing` when returning
BEFORE CODE: @test setindex!(fill(0.0), fill(1.0), CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, CI0) == fill(1.0)
@test setindex!(f... | fix the test
```suggestion
@test Base.infer_return_type(checkbounds, Tuple{typ, Any}) <: Nothing
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
julia> checkbounds(Bool, A, 1:3)
true
julia> checkbounds(Bool, A, 1:3, 2:4)
false
```
"""
function checkbounds(::Type{Bool}, A::AbstractArray, I...)
@inline
checkbounds_indices(Bool, axes(A), I)
end
# Linear indexing is explicitly allowed when there is only one (non-cartesian) index;
# indices that do not all... | Revert doc change.
```suggestion
``` | julia> checkbounds(Bool, A, 1:3)
true
julia> checkbounds(Bool, A, 1:3, 2:4)
false
```
"""
function checkbounds(::Type{Bool}, A::AbstractArray, I...)
@inline
checkbounds_indices(Bool, axes(A), I)
end
# Linear indexing is explicitly allowed when there is only one (non-cartesian) index;
# indices that do not all... | @@ -696,6 +696,8 @@ end
checkbounds(A, I...)
Throw an error if the specified indices `I` are not in bounds for the given array `A`.
+
+In case there is no throw, it is recommended to return [`nothing`](@ref). | base/abstractarray.jl | 26 | Julia | 0.5 | suggestion | 37 | 51 | 51 | false | make `checkbounds(collection, indices...)` return `nothing` when returning | 60,956 | JuliaLang/julia | 35,573 | Julia | nsajko | nsajko | Review this pull request change.
Language: Julia
File: base/abstractarray.jl
PR Title: make `checkbounds(collection, indices...)` return `nothing` when returning
BEFORE CODE: julia> checkbounds(Bool, A, 1:3)
true
julia> checkbounds(Bool, A, 1:3, 2:4)
false
```
"""
function checkbounds(::Type{Bool}, A::AbstractArra... | Revert doc change.
```suggestion
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@test setindex!(fill(0.0), fill(1.0), CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, :) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0, CI0) == fill(1.0)
... | Out of curiosity, why not test something like
```
all(==(...), Base.return_types(checkbounds, Tuple{Any, Any})
```
? | @test setindex!(fill(0.0), fill(1.0), CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, :) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0, CI0) == fill(1.0)
... | @@ -3256,6 +3256,12 @@ end
@test setindex!(zeros(2,2), fill(1.0), CI0, 1, 1) == [1.0 0.0; 0.0 0.0]
end
+@testset "conditionally-throwing version of `checkbounds` should return `nothing` if it returns" begin
+ for typ in (AbstractString, DenseArray, Base.AbstractBroadcasted)
+ @test Base.infer_return_t... | test/arrayops.jl | 26 | Julia | 0.857 | question | 118 | 51 | 51 | false | make `checkbounds(collection, indices...)` return `nothing` when returning | 60,956 | JuliaLang/julia | 35,573 | Julia | KristofferC | nsajko | Review this pull request change.
Language: Julia
File: test/arrayops.jl
PR Title: make `checkbounds(collection, indices...)` return `nothing` when returning
BEFORE CODE: @test setindex!(fill(0.0), fill(1.0), CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, CI0) == fill(1.0)
@test setindex!(f... | Out of curiosity, why not test something like
```
all(==(...), Base.return_types(checkbounds, Tuple{Any, Any})
```
? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@test setindex!(fill(0.0), fill(1.0), CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, :) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0, CI0) == fill(1.0)
... | fix: make the test more strict:
```suggestion
@test Base.infer_return_type(checkbounds, Tuple{typ, Vararg}) <: Nothing
``` | @test setindex!(fill(0.0), fill(1.0), CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, :) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0, CI0) == fill(1.0)
... | @@ -3256,6 +3256,12 @@ end
@test setindex!(zeros(2,2), fill(1.0), CI0, 1, 1) == [1.0 0.0; 0.0 0.0]
end
+@testset "conditionally-throwing version of `checkbounds` should return `nothing` if it returns" begin
+ for typ in (AbstractString, DenseArray, Base.AbstractBroadcasted)
+ @test Base.infer_return_t... | test/arrayops.jl | 26 | Julia | 0.857 | suggestion | 131 | 51 | 51 | false | make `checkbounds(collection, indices...)` return `nothing` when returning | 60,956 | JuliaLang/julia | 35,573 | Julia | nsajko | nsajko | Review this pull request change.
Language: Julia
File: test/arrayops.jl
PR Title: make `checkbounds(collection, indices...)` return `nothing` when returning
BEFORE CODE: @test setindex!(fill(0.0), fill(1.0), CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, CI0) == fill(1.0)
@test setindex!(f... | fix: make the test more strict:
```suggestion
@test Base.infer_return_type(checkbounds, Tuple{typ, Vararg}) <: Nothing
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# 0-dimensional
@test setindex!(fill(0.0), fill(1.0), CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, :) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0,... | make the test even more strict:
```suggestion
for typ in (AbstractString, AbstractArray, Base.AbstractBroadcasted)
``` | # 0-dimensional
@test setindex!(fill(0.0), fill(1.0), CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, :) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0,... | @@ -3256,6 +3256,12 @@ end
@test setindex!(zeros(2,2), fill(1.0), CI0, 1, 1) == [1.0 0.0; 0.0 0.0]
end
+@testset "conditionally-throwing version of `checkbounds` should return `nothing` if it returns" begin
+ for typ in (AbstractString, DenseArray, Base.AbstractBroadcasted) | test/arrayops.jl | 26 | Julia | 0.857 | refactor | 123 | 51 | 51 | false | make `checkbounds(collection, indices...)` return `nothing` when returning | 60,956 | JuliaLang/julia | 35,573 | Julia | nsajko | nsajko | Review this pull request change.
Language: Julia
File: test/arrayops.jl
PR Title: make `checkbounds(collection, indices...)` return `nothing` when returning
BEFORE CODE: # 0-dimensional
@test setindex!(fill(0.0), fill(1.0), CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, CI0) == fill(1.0)
... | make the test even more strict:
```suggestion
for typ in (AbstractString, AbstractArray, Base.AbstractBroadcasted)
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@test setindex!(fill(0.0), fill(1.0), CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, :) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0, CI0) == fill(1.0)
... | ```suggestion
# can not just set `typ = Any` because that would include the predicate (non-throwing) methods of `checkbounds`, because `Type{Bool} <: Any`
for typ in (AbstractString, AbstractArray, Base.AbstractBroadcasted)
``` | @test setindex!(fill(0.0), fill(1.0), CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, :) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), :, CI0, CI0) == fill(1.0)
... | @@ -3256,6 +3256,12 @@ end
@test setindex!(zeros(2,2), fill(1.0), CI0, 1, 1) == [1.0 0.0; 0.0 0.0]
end
+@testset "conditionally-throwing version of `checkbounds` should return `nothing` if it returns" begin
+ for typ in (AbstractString, AbstractArray, Base.AbstractBroadcasted) | test/arrayops.jl | 26 | Julia | 1 | refactor | 235 | 51 | 51 | false | make `checkbounds(collection, indices...)` return `nothing` when returning | 60,956 | JuliaLang/julia | 35,573 | Julia | nsajko | nsajko | Review this pull request change.
Language: Julia
File: test/arrayops.jl
PR Title: make `checkbounds(collection, indices...)` return `nothing` when returning
BEFORE CODE: @test setindex!(fill(0.0), fill(1.0), CI0) == fill(1.0)
@test setindex!(fill(0.0), fill(1.0), CI0, CI0) == fill(1.0)
@test setindex!(f... | ```suggestion
# can not just set `typ = Any` because that would include the predicate (non-throwing) methods of `checkbounds`, because `Type{Bool} <: Any`
for typ in (AbstractString, AbstractArray, Base.AbstractBroadcasted)
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# This file is a part of Julia. License is MIT: https://julialang.org/license
## File Operations (Libuv-based) ##
"""
JL_O_APPEND
JL_O_ASYNC
JL_O_CLOEXEC
JL_O_CREAT
JL_O_DIRECT
JL_O_DIRECTORY
JL_O_DSYNC
JL_O_EXCL
JL_O_FSYNC
JL_O_LARGEFILE
JL_O_NDELAY
JL_O_NOATIME
JL... | ```suggestion
# n.b. this file is `Core.eval`-uated into (pre-existing) Base.Filesystem
# module Filesystem
``` | # This file is a part of Julia. License is MIT: https://julialang.org/license
# NB: This file is `Core.eval`-uated into the (pre-existing) module Filesystem
## File Operations (Libuv-based) ##
"""
JL_O_APPEND
JL_O_ASYNC
JL_O_CLOEXEC
JL_O_CREAT
JL_O_DIRECT
JL_O_DIRECTORY
JL_O_DSYNC
JL_... | @@ -2,8 +2,6 @@
## File Operations (Libuv-based) ##
| base/filesystem.jl | 6 | Julia | 0.714 | suggestion | 111 | 31 | 31 | false | Feature/change bootstrap to use libdl in pcre | 60,967 | JuliaLang/julia | 35,573 | Julia | topolarity | johroj | Review this pull request change.
Language: Julia
File: base/filesystem.jl
PR Title: Feature/change bootstrap to use libdl in pcre
BEFORE CODE: # This file is a part of Julia. License is MIT: https://julialang.org/license
## File Operations (Libuv-based) ##
"""
JL_O_APPEND
JL_O_ASYNC
JL_O_CLOEXEC
J... | ```suggestion
# n.b. this file is `Core.eval`-uated into (pre-existing) Base.Filesystem
# module Filesystem
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
include("iobuffer.jl")
# Concurrency (part 1)
include("linked_list.jl")
include("condition.jl")
include("threads.jl")
include("lock.jl")
# strings & printing
include("intfuncs.jl")
include("strings/strings.jl")
#=
isdebugbuild is defined here as this is imported in libdl.jl (included in libc.jl)
=#
"""
isdebugbu... | Can we define these "multi-stage" modules in `Base.jl` directly and then `Core.eval` into them anywhere they have definitions added?
Makes it easier to see how these modules are built w/o relying on comments to stay up-to-date | include("iobuffer.jl")
# Concurrency (part 1)
include("linked_list.jl")
include("condition.jl")
include("threads.jl")
include("lock.jl")
# strings & printing
include("intfuncs.jl")
include("strings/strings.jl")
#=
isdebugbuild is defined here as this is imported in libdl.jl (included in libc.jl)
=#
"""
isdebugbu... | @@ -106,6 +106,27 @@ include("lock.jl")
# strings & printing
include("intfuncs.jl")
include("strings/strings.jl")
+
+#=
+isdebugbuild is defined here as this is imported in libdl.jl (included in libc.jl)
+=#
+"""
+ isdebugbuild()
+
+Return `true` if julia is a debug version.
+"""
+function isdebugbuild()
+ ret... | base/Base.jl | 26 | Julia | 0.571 | question | 227 | 51 | 51 | false | Feature/change bootstrap to use libdl in pcre | 60,967 | JuliaLang/julia | 35,573 | Julia | topolarity | johroj | Review this pull request change.
Language: Julia
File: base/Base.jl
PR Title: Feature/change bootstrap to use libdl in pcre
BEFORE CODE: include("iobuffer.jl")
# Concurrency (part 1)
include("linked_list.jl")
include("condition.jl")
include("threads.jl")
include("lock.jl")
# strings & printing
include("intfuncs.j... | Can we define these "multi-stage" modules in `Base.jl` directly and then `Core.eval` into them anywhere they have definitions added?
Makes it easier to see how these modules are built w/o relying on comments to stay up-to-date | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# tm.isdst = -1
if r == C_NULL
# TODO: better error message
throw(ArgumentError("invalid arguments"))
end
@static if Sys.isapple()
function shouldcallmktime(s::AbstractString)
# Equivalent to !occursin(r"([^%]|^)%(a|A|j|w|Ow)"a, s), but
# without regex sin... | ```suggestion
i = findnext(==(UInt8('%')), c, i+1)
``` | # tm.isdst = -1
if r == C_NULL
# TODO: better error message
throw(ArgumentError("invalid arguments"))
end
@static if Sys.isapple()
function shouldcallmktime(s::AbstractString)
# Equivalent to !occursin(r"([^%]|^)%(a|A|j|w|Ow)"a, s), but
# without regex sin... | @@ -266,9 +266,32 @@ function strptime(fmt::AbstractString, timestr::AbstractString)
throw(ArgumentError("invalid arguments"))
end
@static if Sys.isapple()
+ function shouldcallmktime(s::AbstractString)
+ # Equivalent to !occursin(r"([^%]|^)%(a|A|j|w|Ow)"a, s), but
+ # wi... | base/libc.jl | 26 | Julia | 0.571 | suggestion | 70 | 51 | 51 | false | Feature/change bootstrap to use libdl in pcre | 60,967 | JuliaLang/julia | 35,573 | Julia | topolarity | johroj | Review this pull request change.
Language: Julia
File: base/libc.jl
PR Title: Feature/change bootstrap to use libdl in pcre
BEFORE CODE: # tm.isdst = -1
if r == C_NULL
# TODO: better error message
throw(ArgumentError("invalid arguments"))
end
@static if Sys.isapple()
function... | ```suggestion
i = findnext(==(UInt8('%')), c, i+1)
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
r = ccall(:strptime, Cstring, (Cstring, Cstring, Ref{TmStruct}), timestr, fmt, tm)
# the following would tell mktime() that this is a local time, and that
# it should try to guess the timezone. not sure if/how this should be
# exposed in the API.
# tm.isdst = -1
if r == C_NULL
# TODO: be... | ```suggestion
if c[i+1] in b"aAjw"
return false
elseif c[i+1] == UInt8('O')
(i <= N-2) && (c[i+2] == UInt8('w')) && return false
```
or similar
(to avoid the `Vector{UInt8}` allocation from the slice) | r = ccall(:strptime, Cstring, (Cstring, Cstring, Ref{TmStruct}), timestr, fmt, tm)
# the following would tell mktime() that this is a local time, and that
# it should try to guess the timezone. not sure if/how this should be
# exposed in the API.
# tm.isdst = -1
if r == C_NULL
# TODO: be... | @@ -266,9 +266,32 @@ function strptime(fmt::AbstractString, timestr::AbstractString)
throw(ArgumentError("invalid arguments"))
end
@static if Sys.isapple()
+ function shouldcallmktime(s::AbstractString)
+ # Equivalent to !occursin(r"([^%]|^)%(a|A|j|w|Ow)"a, s), but
+ # wi... | base/libc.jl | 26 | Julia | 0.857 | suggestion | 289 | 51 | 51 | false | Feature/change bootstrap to use libdl in pcre | 60,967 | JuliaLang/julia | 35,573 | Julia | topolarity | johroj | Review this pull request change.
Language: Julia
File: base/libc.jl
PR Title: Feature/change bootstrap to use libdl in pcre
BEFORE CODE: r = ccall(:strptime, Cstring, (Cstring, Cstring, Ref{TmStruct}), timestr, fmt, tm)
# the following would tell mktime() that this is a local time, and that
# it should ... | ```suggestion
if c[i+1] in b"aAjw"
return false
elseif c[i+1] == UInt8('O')
(i <= N-2) && (c[i+2] == UInt8('w')) && return false
```
or similar
(to avoid the `Vector{UInt8}` allocation from the slice) | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
strptime(timestr::AbstractString) = strptime("%c", timestr)
function strptime(fmt::AbstractString, timestr::AbstractString)
tm = TmStruct()
r = ccall(:strptime, Cstring, (Cstring, Cstring, Ref{TmStruct}), timestr, fmt, tm)
# the following would tell mktime() that this is a local time, and that
# it shou... | ```suggestion
if !(i > firstindex(c) && c[i-1] == UInt8('%'))
``` | strptime(timestr::AbstractString) = strptime("%c", timestr)
function strptime(fmt::AbstractString, timestr::AbstractString)
tm = TmStruct()
r = ccall(:strptime, Cstring, (Cstring, Cstring, Ref{TmStruct}), timestr, fmt, tm)
# the following would tell mktime() that this is a local time, and that
# it shou... | @@ -266,9 +266,32 @@ function strptime(fmt::AbstractString, timestr::AbstractString)
throw(ArgumentError("invalid arguments"))
end
@static if Sys.isapple()
+ function shouldcallmktime(s::AbstractString)
+ # Equivalent to !occursin(r"([^%]|^)%(a|A|j|w|Ow)"a, s), but
+ # wi... | base/libc.jl | 26 | Julia | 0.571 | suggestion | 81 | 51 | 51 | false | Feature/change bootstrap to use libdl in pcre | 60,967 | JuliaLang/julia | 35,573 | Julia | topolarity | johroj | Review this pull request change.
Language: Julia
File: base/libc.jl
PR Title: Feature/change bootstrap to use libdl in pcre
BEFORE CODE: strptime(timestr::AbstractString) = strptime("%c", timestr)
function strptime(fmt::AbstractString, timestr::AbstractString)
tm = TmStruct()
r = ccall(:strptime, Cstring, (... | ```suggestion
if !(i > firstindex(c) && c[i-1] == UInt8('%'))
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
vb.constraintkind = vb.concrete ? 1 : 2;
else if (u->var->lb != jl_bottom_type)
vb.constraintkind = 2;
else if (vb.occurs_cov && noinv)
vb.constraintkind = 1;
int reintersection = constraint1 != vb.constraintkind || vb.concrete;
if (reintersection) {
... | I'm suspicious of a call to `jl_subtype` happening from this deep inside the subtyping algorithm, ignoring all the current status of the algorithm | else if (jl_is_uniontype(v)) {
return always_occurs_cov(((jl_uniontype_t*)v)->a, var, param) &&
always_occurs_cov(((jl_uniontype_t*)v)->b, var, param);
}
else if (jl_is_unionall(v)) {
jl_unionall_t *ua = (jl_unionall_t*)v;
return ua->var != var && (
always_... | @@ -3497,6 +3521,34 @@ static jl_value_t *intersect_unionall(jl_value_t *t, jl_unionall_t *u, jl_stenv_
res = intersect_unionall_(t, u, e, R, param, &vb);
}
}
+ else if (res == jl_bottom_type && vb.constraintkind == 1 &&
+ vb.occurs_cov > 1) {
+ // vb.constraintkind == 1... | src/subtype.c | 26 | C | 0.643 | suggestion | 146 | 51 | 51 | false | fix intersection returning Union{} for diagonal vars with different Type{X} arguments | 60,989 | JuliaLang/julia | 35,573 | Julia | vtjnash | adienes | Review this pull request change.
Language: C
File: src/subtype.c
PR Title: fix intersection returning Union{} for diagonal vars with different Type{X} arguments
BEFORE CODE: vb.constraintkind = vb.concrete ? 1 : 2;
else if (u->var->lb != jl_bottom_type)
vb.constraintkind = 2;
... | I'm suspicious of a call to `jl_subtype` happening from this deep inside the subtyping algorithm, ignoring all the current status of the algorithm | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
{
return is_leaf_bound(v->lb);
}
static jl_value_t *widen_Type(jl_value_t *t JL_PROPAGATES_ROOT) JL_NOTSAFEPOINT
{
if (jl_is_type_type(t) && !jl_is_typevar(jl_tparam0(t)))
return jl_typeof(jl_tparam0(t));
if (jl_is_uniontype(t)) {
jl_value_t *a = widen_Type(((jl_uniontype_t*)t)->a);
... | Perhaps we can fuse this with the `widen_Type` by adding a `concrete_only` like flag?
```c
static jl_value_t *widen_Type(jl_value_t *t JL_PROPAGATES_ROOT, int concrete_only) JL_NOTSAFEPOINT
{
if (jl_is_type_type(t) && !jl_is_typevar(jl_tparam0(t)))
return jl_typeof(jl_tparam0(t));
if (jl_is_uniontype(t... | {
return is_leaf_bound(v->lb);
}
static jl_value_t *widen_Type_if_concrete(jl_value_t *t JL_PROPAGATES_ROOT) JL_NOTSAFEPOINT
{
if (jl_is_type_type(t) && !jl_is_typevar(jl_tparam0(t)))
return jl_typeof(jl_tparam0(t));
if (jl_is_uniontype(t)) {
jl_value_t *a = widen_Type_if_concrete(((jl_unio... | @@ -886,6 +887,27 @@ static jl_value_t *widen_Type(jl_value_t *t JL_PROPAGATES_ROOT) JL_NOTSAFEPOINT
return t;
}
+// Map Type{X} to kind type (DataType, UnionAll, Union, TypeofBottom) over union | src/subtype.c | 26 | C | 1 | suggestion | 1,496 | 51 | 51 | false | fix intersection returning Union{} for diagonal vars with different Type{X} arguments | 60,989 | JuliaLang/julia | 35,573 | Julia | N5N3 | adienes | Review this pull request change.
Language: C
File: src/subtype.c
PR Title: fix intersection returning Union{} for diagonal vars with different Type{X} arguments
BEFORE CODE: {
return is_leaf_bound(v->lb);
}
static jl_value_t *widen_Type(jl_value_t *t JL_PROPAGATES_ROOT) JL_NOTSAFEPOINT
{
if (jl_is_type_typ... | Perhaps we can fuse this with the `widen_Type` by adding a `concrete_only` like flag?
```c
static jl_value_t *widen_Type(jl_value_t *t JL_PROPAGATES_ROOT, int concrete_only) JL_NOTSAFEPOINT
{
if (jl_is_type_type(t) && !jl_is_typevar(jl_tparam0(t)))
return jl_typeof(jl_tparam0(t));
if (jl_is_uniontype(t... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
vb.intersected = 1;
if (vb.limited) {
// if the environment got too big, avoid tree recursion and propagate the flag
if (e->vars)
e->vars->limited = 1;
}
else if (res != jl_bottom_type) {
int constraint1 = vb.constraintkind;
if (vb.concrete || vb.occurs_inv>1 ... | With the change below (always try Type widening for constraintkind == 1), we can trust the intersection result (no false `Union{}`), and only trigger the final re-intersection if `widen_to_kind == 1` && `occurs_cov > 1` && `vb.ub is a kind`
```c
if (res != jl_bottom_type && vb.constraintkind == 1 && vb.widen_to_k... |
static int always_occurs_cov(jl_value_t *v, jl_tvar_t *var, int param) JL_NOTSAFEPOINT
{
if (param > 1) {
return 0;
}
else if (v == (jl_value_t*)var) {
return param == 1;
}
else if (jl_is_uniontype(v)) {
return always_occurs_cov(((jl_uniontype_t*)v)->a, var, param) &&
... | @@ -3497,6 +3521,37 @@ static jl_value_t *intersect_unionall(jl_value_t *t, jl_unionall_t *u, jl_stenv_
res = intersect_unionall_(t, u, e, R, param, &vb);
}
}
+ else if (res == jl_bottom_type && vb.constraintkind == 1 && | src/subtype.c | 26 | C | 1 | suggestion | 549 | 51 | 51 | false | fix intersection returning Union{} for diagonal vars with different Type{X} arguments | 60,989 | JuliaLang/julia | 35,573 | Julia | N5N3 | adienes | Review this pull request change.
Language: C
File: src/subtype.c
PR Title: fix intersection returning Union{} for diagonal vars with different Type{X} arguments
BEFORE CODE: vb.intersected = 1;
if (vb.limited) {
// if the environment got too big, avoid tree recursion and propagate the flag
i... | With the change below (always try Type widening for constraintkind == 1), we can trust the intersection result (no false `Union{}`), and only trigger the final re-intersection if `widen_to_kind == 1` && `occurs_cov > 1` && `vb.ub is a kind`
```c
if (res != jl_bottom_type && vb.constraintkind == 1 && vb.widen_to_k... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
if (ub != (jl_value_t*)b) {
if (jl_has_free_typevars(ub)) {
if (check_unsat_bound(ub, b, e)) {
JL_GC_POP();
return jl_bottom_type;
}
}
bb->ub = ub;
if ((jl_is_uniontype(ub) && !jl_is_uniontype... | Perhaps we can always try to widening `bb->ub` if `bb->constraintkind == 1` and let the flag to record if `ub` really got widened, i.e.
```c
if (e->triangular && param == 1) {
if (check_unsat_bound(ub, b, e))
return jl_bottom_type;
set_bound(&bb->ub, ub, b, e);
return (jl_value_t... | if (R) flip_offset(e);
int ccheck = intersect_var_ccheck_in_env(bb->lb, bb->ub, a, a, e, !R);
if (R) flip_offset(e);
if (!ccheck) {
JL_GC_POP();
return jl_bottom_type;
}
ub = a;
}
else {
e... | @@ -2841,6 +2863,8 @@ static jl_value_t *intersect_var(jl_tvar_t *b, jl_value_t *a, jl_stenv_t *e, int
if (e->triangular && check_unsat_bound(ub, b, e))
return jl_bottom_type;
set_bound(&bb->ub, ub, b, e);
+ if (bb->constraintkind == 1 && bb->widen_to_kind)
+ set_bound(&... | src/subtype.c | 26 | C | 1 | suggestion | 1,251 | 51 | 51 | false | fix intersection returning Union{} for diagonal vars with different Type{X} arguments | 60,989 | JuliaLang/julia | 35,573 | Julia | N5N3 | adienes | Review this pull request change.
Language: C
File: src/subtype.c
PR Title: fix intersection returning Union{} for diagonal vars with different Type{X} arguments
BEFORE CODE: if (ub != (jl_value_t*)b) {
if (jl_has_free_typevars(ub)) {
if (check_unsat_bound(ub, b, e)) {
... | Perhaps we can always try to widening `bb->ub` if `bb->constraintkind == 1` and let the flag to record if `ub` really got widened, i.e.
```c
if (e->triangular && param == 1) {
if (check_unsat_bound(ub, b, e))
return jl_bottom_type;
set_bound(&bb->ub, ub, b, e);
return (jl_value_t... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
e->invdepth, NULL, e->vars };
JL_GC_PUSH4(&res, &vb.lb, &vb.ub, &vb.innervars);
save_env(e, &se, 1);
int noinv = !var_occurs_invariant(u->body, u->var);
if (is_leaf_typevar(u->var) && noinv && always_occurs_cov(u->body, u->var, param))
vb.constraintkind = 1;
res = ... | In fact, non-leaf `lb` (rather than `ub`) violates the diagonal rule. Looks like the remaining `@test_broken` stems from this. | }
else if (res != jl_bottom_type) {
int constraint1 = vb.constraintkind;
if (vb.concrete || vb.occurs_inv>1 || (vb.occurs_inv && vb.occurs_cov))
vb.constraintkind = vb.concrete ? 1 : 2;
else if (u->var->lb != jl_bottom_type)
vb.constraintkind = 2;
else if ... | @@ -3533,35 +3578,30 @@ static jl_value_t *intersect_unionall(jl_value_t *t, jl_unionall_t *u, jl_stenv_
res = intersect_unionall_(t, u, e, R, param, &vb);
}
}
- else if (res == jl_bottom_type && vb.constraintkind == 1 &&
- vb.occurs_cov > 1) {
- // vb.constraintkind == ... | src/subtype.c | 26 | C | 0.5 | bug | 126 | 51 | 51 | false | fix intersection returning Union{} for diagonal vars with different Type{X} arguments | 60,989 | JuliaLang/julia | 35,573 | Julia | N5N3 | adienes | Review this pull request change.
Language: C
File: src/subtype.c
PR Title: fix intersection returning Union{} for diagonal vars with different Type{X} arguments
BEFORE CODE: e->invdepth, NULL, e->vars };
JL_GC_PUSH4(&res, &vb.lb, &vb.ub, &vb.innervars);
save_env(e, &se, 1);
in... | In fact, non-leaf `lb` (rather than `ub`) violates the diagonal rule. Looks like the remaining `@test_broken` stems from this. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
jl_value_t *a = widen_Type_if_concrete(((jl_uniontype_t*)t)->a);
jl_value_t *b = widen_Type_if_concrete(((jl_uniontype_t*)t)->b);
if (a == b)
return a;
}
if (jl_is_unionall(t)) {
jl_unionall_t *u = (jl_unionall_t*)t;
jl_value_t *body = widen_Type_if_concrete(u... | I'm not sure return a `Union{}` is a good idea here, this might remove part of `ub` and make the result too narrow, e.g.
```julia
typeintersect(Tuple{Union{Type{Val},Type{Int8}},Union{Type{Val},Type{Int}}}, NTuple{2,<:Union{Type{<:Val},DataType}})
```
why not just return `t`?
BTW, there seems no need to maintain `filt... | jl_value_t *a = widen_Type_if_concrete(((jl_uniontype_t*)t)->a);
jl_value_t *b = widen_Type_if_concrete(((jl_uniontype_t*)t)->b);
if (a == b)
return a;
}
if (jl_is_unionall(t)) {
jl_unionall_t *u = (jl_unionall_t*)t;
jl_value_t *body = widen_Type_if_concrete(u... | @@ -874,49 +874,70 @@ static int is_leaf_typevar(jl_tvar_t *v) JL_NOTSAFEPOINT
return is_leaf_bound(v->lb);
}
-static jl_value_t *widen_Type(jl_value_t *t JL_PROPAGATES_ROOT) JL_NOTSAFEPOINT
+static jl_value_t *widen_Type_if_concrete(jl_value_t *t JL_PROPAGATES_ROOT) JL_NOTSAFEPOINT
{
if (jl_is_type_type(... | src/subtype.c | 26 | C | 1 | suggestion | 492 | 51 | 51 | false | fix intersection returning Union{} for diagonal vars with different Type{X} arguments | 60,989 | JuliaLang/julia | 35,573 | Julia | N5N3 | adienes | Review this pull request change.
Language: C
File: src/subtype.c
PR Title: fix intersection returning Union{} for diagonal vars with different Type{X} arguments
BEFORE CODE: jl_value_t *a = widen_Type_if_concrete(((jl_uniontype_t*)t)->a);
jl_value_t *b = widen_Type_if_concrete(((jl_uniontype_t*)t)->... | I'm not sure return a `Union{}` is a good idea here, this might remove part of `ub` and make the result too narrow, e.g.
```julia
typeintersect(Tuple{Union{Type{Val},Type{Int8}},Union{Type{Val},Type{Int}}}, NTuple{2,<:Union{Type{<:Val},DataType}})
```
why not just return `t`?
BTW, there seems no need to maintain `filt... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
push!(method_stmts,
method_def_expr(ctx, srcref, callex, method_table,
trimmed_typevar_names, trimmed_arg_names, trimmed_arg_types,
body))
end
end
function scope_nest(ctx, names, values, body)
for (name, value) in Iterators.rever... | Does this comment still apply? | push!(method_stmts,
method_def_expr(ctx, srcref, callex, method_table,
trimmed_typevar_names, trimmed_arg_names, trimmed_arg_types,
body))
end
end
function scope_nest(ctx, names, values, body)
for (name, value) in Iterators.rever... | @@ -2614,13 +2614,15 @@ function keyword_function_defs(ctx, srcref, callex_srcref, name_str, typevar_nam
end
# TODO: Is the layer correct here? Which module should be the parent module
# of this body function? | JuliaLowering/src/desugaring.jl | 26 | Julia | 0.071 | question | 30 | 51 | 51 | false | [JuliaLowering] Fix closure-conversion of keyword functions | 61,001 | JuliaLang/julia | 35,573 | Julia | topolarity | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/desugaring.jl
PR Title: [JuliaLowering] Fix closure-conversion of keyword functions
BEFORE CODE: push!(method_stmts,
method_def_expr(ctx, srcref, callex, method_table,
trimmed_typevar_names, ... | Does this comment still apply? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
push!(kwcall_method_defs,
method_def_expr(ctx, srcref, callex_srcref, kwcall_mtable,
main_kwcall_typevars, kwcall_arg_names, kwcall_arg_types, kwcall_body))
# Check kws of body method
check_all_typevars_used(body_arg_types, typevar_names, typevar_stmts)
kw_func_meth... | ```suggestion
# This should inherit the local / global status of the body func,
# so provide that here for closure conversion, even though this is
# really a method on Core.kwcall
body_func_name
```
I assume this is done for closure conversion, but the comment seems to ... | main_kwcall_typevars, kwcall_arg_names, kwcall_arg_types, kwcall_body))
# Check kws of body method
check_all_typevars_used(body_arg_types, typevar_names, typevar_stmts)
kw_func_method_defs = @ast ctx srcref [K"block"
[K"function_decl" body_func_name]
[K"method_def... | @@ -2860,7 +2862,9 @@ function keyword_function_defs(ctx, srcref, callex_srcref, name_str, typevar_nam
]
]
[K"method_defs"
- "nothing"::K"core"
+ # We want this method to be local or global at the same time as the
+ # body func, but it's really a method on... | JuliaLowering/src/desugaring.jl | 26 | Julia | 0.857 | suggestion | 556 | 51 | 51 | false | [JuliaLowering] Fix closure-conversion of keyword functions | 61,001 | JuliaLang/julia | 35,573 | Julia | topolarity | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/desugaring.jl
PR Title: [JuliaLowering] Fix closure-conversion of keyword functions
BEFORE CODE: push!(kwcall_method_defs,
method_def_expr(ctx, srcref, callex_srcref, kwcall_mtable,
main_kwcall_typevars,... | ```suggestion
# This should inherit the local / global status of the body func,
# so provide that here for closure conversion, even though this is
# really a method on Core.kwcall
body_func_name
```
I assume this is done for closure conversion, but the comment seems to ... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
function _match_kind(srcref, ex)
kws = []
if Meta.isexpr(ex, :call)
kind = esc(ex.args[1])
args = ex.args[2:end]
if Meta.isexpr(args[1], :parameters)
kws = map(_kw_to_pair, args[1].args)
popfirst!(args)
end
while length(args) >= 1 && Meta.isexpr(a... | `line` makes me think this has semantic (not debug) meaning |
function _match_kind(srcref, ex)
kws = []
if Meta.isexpr(ex, :call)
kind = esc(ex.args[1])
args = ex.args[2:end]
if Meta.isexpr(args[1], :parameters)
kws = map(_kw_to_pair, args[1].args)
popfirst!(args)
end
while length(args) >= 1 && Meta.isexpr(a... | @@ -233,7 +233,7 @@ function _match_kind(srcref, ex)
return (kind, srcref, kws)
end
-function _expand_ast_tree(ctx, srcref, tree)
+function _expand_ast_tree(ctx, srcref, tree, line::QuoteNode) | JuliaLowering/src/ast.jl | 26 | Julia | 0.5 | bug | 59 | 51 | 51 | false | SyntaxTree: Add `jl_source` attribute for debugging JuliaLowering | 61,006 | JuliaLang/julia | 35,573 | Julia | topolarity | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/ast.jl
PR Title: SyntaxTree: Add `jl_source` attribute for debugging JuliaLowering
BEFORE CODE:
function _match_kind(srcref, ex)
kws = []
if Meta.isexpr(ex, :call)
kind = esc(ex.args[1])
args = ex.args[2:end]
i... | `line` makes me think this has semantic (not debug) meaning | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# endif
#else
# ifdef JL_NDEBUG
# define NDEBUG
# include <assert.h>
# undef NDEBUG
# else
# include <assert.h>
# endif
#endif
#if defined __has_builtin
# define jl_has_builtin(x) __has_builtin(x)
#else
# define jl_has_builtin(x) 0
#endif
#ifdef JL_NDEBUG
# if jl_has_builtin(__builtin_assume)
#defin... | the space-indenting for these directives seems inconsistent | // This file is a part of Julia. License is MIT: https://julialang.org/license
// Include this file instead of `assert.h` directly.
// This is necessary because LLVM sometimes has bugs that cause runtime assertion if
// the `NDEBUG` setting is different from the one used to compile LLVM.
// For C++ files, we set `NDEB... | @@ -29,4 +35,100 @@
# include <assert.h>
# endif
#endif
+
+#if defined __has_builtin
+# define jl_has_builtin(x) __has_builtin(x)
+#else
+# define jl_has_builtin(x) 0
+#endif
+
+#ifdef JL_NDEBUG
+# if jl_has_builtin(__builtin_assume)
+#define jl_assume(cond) \
+ (__extension__({ ... | src/julia_assert.h | 26 | C/C++ | 0.214 | style | 59 | 51 | 33 | false | Make jl_assume, jl_assume_aligned, jl_unreachable asserts in debug | 61,015 | JuliaLang/julia | 35,573 | Julia | topolarity | xal-0 | Review this pull request change.
Language: C/C++
File: src/julia_assert.h
PR Title: Make jl_assume, jl_assume_aligned, jl_unreachable asserts in debug
BEFORE CODE: # endif
#else
# ifdef JL_NDEBUG
# define NDEBUG
# include <assert.h>
# undef NDEBUG
# else
# include <assert.h>
# endif
#endif
#if defi... | the space-indenting for these directives seems inconsistent | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
#define FORCE_INLINE static inline __attribute__((always_inline))
#define EXTERN_INLINE_DECLARE inline __attribute__ ((visibility("default")))
#define EXTERN_INLINE_DEFINE extern inline JL_DLLEXPORT
#if defined(_OS_WINDOWS_) && !defined(_COMPILER_GCC_)
# define NOINLINE __declspec(noinline)
# define NOINLINE_DECL(f... | ```suggestion
#if defined(__has_builtin)
```
just a tiny style nit, since we (mostly) use this form | #define FORCE_INLINE static inline __attribute__((always_inline))
#define EXTERN_INLINE_DECLARE inline __attribute__ ((visibility("default")))
#define EXTERN_INLINE_DEFINE extern inline JL_DLLEXPORT
#if defined(_OS_WINDOWS_) && !defined(_COMPILER_GCC_)
# define NOINLINE __declspec(noinline)
# define NOINLINE_DECL(f... | @@ -142,52 +143,87 @@ typedef intptr_t ssize_t;
# define JL_ATTRIBUTE_ALIGN_PTRSIZE(x)
#endif
-#ifdef __has_builtin
+#if defined __has_builtin | src/support/dtypes.h | 26 | C/C++ | 0.786 | suggestion | 100 | 51 | 51 | false | Make jl_assume, jl_assume_aligned, jl_unreachable asserts in debug | 61,015 | JuliaLang/julia | 35,573 | Julia | topolarity | xal-0 | Review this pull request change.
Language: C/C++
File: src/support/dtypes.h
PR Title: Make jl_assume, jl_assume_aligned, jl_unreachable asserts in debug
BEFORE CODE: #define FORCE_INLINE static inline __attribute__((always_inline))
#define EXTERN_INLINE_DECLARE inline __attribute__ ((visibility("default")))
#defin... | ```suggestion
#if defined(__has_builtin)
```
just a tiny style nit, since we (mostly) use this form | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# Desugaring uses this helper internally, so we may see K"kw" too.
if k == K"=" && numchildren(ex) === 2
ex = @ast ctx ex [K"kw" ex[1] ex[2]]
end
mapchildren(c->_apply_nospecialize(ctx, c), ctx, ex, 1:1)
else
throw(LoweringError(ex, "Invalid function argument"))
... | ```suggestion
function Base.var"@nospecialize"(__context__::MacroContext, exs::SyntaxTree...)
if length(exs) == 0
@ast __context__ __context__.macrocall [K"meta" "nospecialize"::K"Symbol"]
elseif length(exs) == 1
_apply_nospecialize(__context__, only(exs))
else
@ast __context__ __con... | # Desugaring uses this helper internally, so we may see K"kw" too.
if k == K"=" && numchildren(ex) === 2
ex = @ast ctx ex [K"kw" ex[1] ex[2]]
end
mapchildren(c->_apply_nospecialize(ctx, c), ctx, ex, 1:1)
else
throw(LoweringError(ex, "Invalid function argument"))
... | @@ -32,11 +32,23 @@ function _apply_nospecialize(ctx, ex)
end
end
-function Base.var"@nospecialize"(__context__::MacroContext, ex, exs...)
- # TODO support multi-arg version properly
+function Base.var"@nospecialize"(__context__::MacroContext)
+ @ast __context__ __context__.macrocall [K"meta" "nospecializ... | JuliaLowering/src/syntax_macros.jl | 26 | Julia | 0.786 | suggestion | 516 | 51 | 51 | false | [JuliaLowering] Fix `@nospecialize` multi-arg and add 0-arg | 61,021 | JuliaLang/julia | 35,573 | Julia | topolarity | aviatesk | Review this pull request change.
Language: Julia
File: JuliaLowering/src/syntax_macros.jl
PR Title: [JuliaLowering] Fix `@nospecialize` multi-arg and add 0-arg
BEFORE CODE: # Desugaring uses this helper internally, so we may see K"kw" too.
if k == K"=" && numchildren(ex) === 2
ex = @ast ... | ```suggestion
function Base.var"@nospecialize"(__context__::MacroContext, exs::SyntaxTree...)
if length(exs) == 0
@ast __context__ __context__.macrocall [K"meta" "nospecialize"::K"Symbol"]
elseif length(exs) == 1
_apply_nospecialize(__context__, only(exs))
else
@ast __context__ __con... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
else
declare_in_scope!(ctx, scope, ex, :local)
end
elseif b.kind === :static_parameter
throw(LoweringError(ex, "cannot overwrite a static parameter"))
elseif b.kind === :local || b.kind === :argument
# unambiguous assignment to existing var... | Writing tests for this PR, I found body-level `@nospecialize` effects haven't been working properly, so I made this change to propagate the effects to `BindingInfo` during scope analysis.
(N.B. argument-level `@nospecialize` effects are processed during desugaring i.e. `_resolve_function_arg`) | # Add local decls to start of block so that closure conversion can
# initialize if necessary.
for id in sort!(collect(values(scope.vars)))
binfo = get_binding(ctx, id)
if binfo.kind == :local
push!(stmts, @ast ctx srcref [K"local" binding_ex(ctx, id)])
end
end
end
fu... | @@ -364,6 +364,11 @@ function _resolve_scopes(ctx, ex::SyntaxTree,
gid = declare_in_scope!(ctx, top_scope(ctx), ex, :global)
b = get_binding(ctx, gid)
end
+ # Body-level @nospecialize sets :nospecialize metadata on identifiers.
+ # Propagate this to the binding so the sl... | JuliaLowering/src/scope_analysis.jl | 26 | Julia | 0.643 | suggestion | 294 | 51 | 51 | false | [JuliaLowering] Fix `@nospecialize` multi-arg and add 0-arg | 61,021 | JuliaLang/julia | 35,573 | Julia | aviatesk | aviatesk | Review this pull request change.
Language: Julia
File: JuliaLowering/src/scope_analysis.jl
PR Title: [JuliaLowering] Fix `@nospecialize` multi-arg and add 0-arg
BEFORE CODE: else
declare_in_scope!(ctx, scope, ex, :local)
end
elseif b.kind === :static_parameter
... | Writing tests for this PR, I found body-level `@nospecialize` effects haven't been working properly, so I made this change to propagate the effects to `BindingInfo` during scope analysis.
(N.B. argument-level `@nospecialize` effects are processed during desugaring i.e. `_resolve_function_arg`) | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
(a, b)
end
f_nospecialize_single_body(1, 2)
end
""") == (1, 2)
@test only(methods(test_mod.f_nospecialize_single_body)).nospecialize == 0b10
# @nospecialize with zero args in function body (blanket nospecialize)
@test JuliaLowering.include_string(test_mod, """
begin... | Would be good to add a broken test for the body form (which doesn't work, as noted in https://github.com/JuliaLang/julia/blob/3a3289a17f5a36feddf0898b67788b894a5c7440/JuliaLowering/src/desugaring.jl#L2582).
A `kwargs` example would also be good (and is similarly broken for the body form). Sorry to make you add so many... | (a, b)
end
f_nospecialize_single_body(1, 2)
end
""") == (1, 2)
@test only(methods(test_mod.f_nospecialize_single_body)).nospecialize == 0b10
# @nospecialize with zero args in function body (blanket nospecialize)
@test JuliaLowering.include_string(test_mod, """
begin... | @@ -378,6 +378,61 @@ end
(f_branch_meta(10, false), f_branch_meta(20, true))
end
""") == (12, 21)
+
+ # @nospecialize with multiple args in function body
+ @test JuliaLowering.include_string(test_mod, """
+ begin
+ function f_nospecialize_multi_body(a, b, c, d)
+ @nospecial... | JuliaLowering/test/functions.jl | 26 | Julia | 0.571 | bug | 388 | 51 | 51 | false | [JuliaLowering] Fix `@nospecialize` multi-arg and add 0-arg | 61,021 | JuliaLang/julia | 35,573 | Julia | topolarity | aviatesk | Review this pull request change.
Language: Julia
File: JuliaLowering/test/functions.jl
PR Title: [JuliaLowering] Fix `@nospecialize` multi-arg and add 0-arg
BEFORE CODE: (a, b)
end
f_nospecialize_single_body(1, 2)
end
""") == (1, 2)
@test only(methods(test_mod.f_nospecialize... | Would be good to add a broken test for the body form (which doesn't work, as noted in https://github.com/JuliaLang/julia/blob/3a3289a17f5a36feddf0898b67788b894a5c7440/JuliaLowering/src/desugaring.jl#L2582).
A `kwargs` example would also be good (and is similarly broken for the body form). Sorry to make you add so many... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
@test JuliaLowering.include_string(test_mod, """
begin
function f_body_nospecialize_default(x, y=1)
@nospecialize
(x, y)
end
(f_body_nospecialize_default(10, 20), f_body_nospecialize_default(30))
end
""") == ((10, 20), (30, 1))
# The 2-arg method has n... | ```suggestion
@test_broken only(methods(Core.kwcall, (NamedTuple,typeof(test_mod.f_body_nospecialize_with_kwargs),Any))).nospecialize == 1 << 2
``` | @test JuliaLowering.include_string(test_mod, """
begin
function f_body_nospecialize_default(x, y=1)
@nospecialize
(x, y)
end
(f_body_nospecialize_default(10, 20), f_body_nospecialize_default(30))
end
""") == ((10, 20), (30, 1))
# The 2-arg method has n... | @@ -378,6 +378,90 @@ end
(f_branch_meta(10, false), f_branch_meta(20, true))
end
""") == (12, 21)
+
+ # @nospecialize with multiple args in function body
+ @test JuliaLowering.include_string(test_mod, """
+ begin
+ function f_nospecialize_multi_body(a, b, c, d)
+ @nospecial... | JuliaLowering/test/functions.jl | 26 | Julia | 0.857 | bug | 151 | 51 | 51 | false | [JuliaLowering] Fix `@nospecialize` multi-arg and add 0-arg | 61,021 | JuliaLang/julia | 35,573 | Julia | topolarity | aviatesk | Review this pull request change.
Language: Julia
File: JuliaLowering/test/functions.jl
PR Title: [JuliaLowering] Fix `@nospecialize` multi-arg and add 0-arg
BEFORE CODE: @test JuliaLowering.include_string(test_mod, """
begin
function f_body_nospecialize_default(x, y=1)
@nospecialize
... | ```suggestion
@test_broken only(methods(Core.kwcall, (NamedTuple,typeof(test_mod.f_body_nospecialize_with_kwargs),Any))).nospecialize == 1 << 2
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
struct ValidationDiagnostic
sts::SyntaxList
msg::String
loc::LineNumberNode # for noting where failures come from in this file
end
ValidationDiagnostic(st::SyntaxTree, msg, loc) =
ValidationDiagnostic(
SyntaxList(syntax_graph(st), NodeId[st._id]), msg, loc)
"""
The type returned by all `vst` fu... | It may be nice to also check for `allowed_attrs`, not just required ?
Assuming that we don't have a good path to make those the same | function assert_syntaxtree(st::SyntaxTree)
vr = _assert_syntaxtree(st, NodeId[], pass())
@assert is_known(vr)
if !vr.ok
msg = string("assert_syntaxtree failed: ", node_string(st), "\n")
for err in vr.errors
msg *= "node: " * node_string(only(err.sts)) * "\nreason: " * err.msg
... | @@ -1027,3 +1028,75 @@ vst0_quoted(vcx, st; quote_level) = @stm st begin
vst0_quoted(vcx, x; quote_level=quote_level+1)
_ -> all(vst0_quoted, vcx, children(st); quote_level)
end
+
+#-------------------------------------------------------------------------------
+# Tree invariants assumed everywhere, incl... | JuliaLowering/src/validation.jl | 26 | Julia | 0.5 | suggestion | 133 | 1,030 | 51 | false | [JuliaLowering] Printing improvements: malformed trees, leaf kinds | 61,028 | JuliaLang/julia | 35,573 | Julia | topolarity | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/validation.jl
PR Title: [JuliaLowering] Printing improvements: malformed trees, leaf kinds
BEFORE CODE: struct ValidationDiagnostic
sts::SyntaxList
msg::String
loc::LineNumberNode # for noting where failures come from in this file
... | It may be nice to also check for `allowed_attrs`, not just required ?
Assuming that we don't have a good path to make those the same | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
n = length(groups)
for i = 1:n
argᵢ = fargs[i]
if argᵢ isa SlotNumber || argᵢ isa SSAValue
for j in 1:i-1
if fargs[j] === argᵢ
groups[i] = groups[j]
break
end
end
end
end
return groups... | When you guys get older and have as bad eyes as I do now, you'll seriously regret using identifiers like tᵢ and tⱼ for being virtually indistinguishable unless you zoom in a lot. (I also don't get what these offer over `ti` or `t_i` -- but you do you, I am just a grumpy old man on the sidelines ;-) ) | for i = 1:length(groups)
arg_i = fargs[i]
if arg_i isa SlotNumber || arg_i isa SSAValue
for j in 1:i-1
if fargs[j] === arg_i
groups[i] = groups[j]
break
end
end
end
end
return groups
end
... | @@ -192,16 +192,90 @@ function tuple_tail_elem(𝕃::AbstractLattice, @nospecialize(init), ct::Vector{
return Vararg{widenconst(t)}
end
+# Given `fargs` from `ArgInfo` and optionally `argtypes`, compute alias groups
+# for argument positions. Returns `nothing` if `fargs === nothing`, otherwise a
+# `Vector{Int}` ... | Compiler/src/typeutils.jl | 26 | Julia | 0.571 | suggestion | 301 | 51 | 51 | false | inference: Alias-aware union splitting for identical arguments | 61,041 | JuliaLang/julia | 35,573 | Julia | fingolfin | aviatesk | Review this pull request change.
Language: Julia
File: Compiler/src/typeutils.jl
PR Title: inference: Alias-aware union splitting for identical arguments
BEFORE CODE: n = length(groups)
for i = 1:n
argᵢ = fargs[i]
if argᵢ isa SlotNumber || argᵢ isa SSAValue
for j in 1:i-1
... | When you guys get older and have as bad eyes as I do now, you'll seriously regret using identifiers like tᵢ and tⱼ for being virtually indistinguishable unless you zoom in a lot. (I also don't get what these offer over `ti` or `t_i` -- but you do you, I am just a grumpy old man on the sidelines ;-) ) | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
na = length(argtypes)
groups = compute_alias_groups(na, fargs, argtypes)
return _switchtupleunion(𝕃, argtypes, na, [], nothing, groups)
end
function _switchtupleunion(𝕃::AbstractLattice, t::Vector{Any}, i::Int, tunion::Vector{Any},
@nospecialize(origt), groups::Union{Nothing,Ve... | nit: The implementation right now constrains that `i > j`, but nothing really says that it must be true>
```suggestion
for j in 1:length(t)
```
Additionally, when we have alias groups, it might be cool to similarly refine dependent types during `smerge!` also | na = length(argtypes)
groups = compute_alias_groups(na, fargs, argtypes)
return _switchtupleunion(𝕃, argtypes, na, [], nothing, groups)
end
function _switchtupleunion(𝕃::AbstractLattice, t::Vector{Any}, i::Int, tunion::Vector{Any},
@nospecialize(origt), groups::Union{Nothing,Ve... | @@ -225,34 +297,69 @@ function switchtupleunion(@nospecialize(ty))
return _switchtupleunion(JLTypeLattice(), Any[tparams...], length(tparams), [], ty)
end
-switchtupleunion(𝕃::AbstractLattice, argtypes::Vector{Any}) = _switchtupleunion(𝕃, argtypes, length(argtypes), [], nothing)
+function switchtupleunion(𝕃:... | Compiler/src/typeutils.jl | 26 | Julia | 0.929 | suggestion | 272 | 51 | 51 | false | inference: Alias-aware union splitting for identical arguments | 61,041 | JuliaLang/julia | 35,573 | Julia | vtjnash | aviatesk | Review this pull request change.
Language: Julia
File: Compiler/src/typeutils.jl
PR Title: inference: Alias-aware union splitting for identical arguments
BEFORE CODE: na = length(argtypes)
groups = compute_alias_groups(na, fargs, argtypes)
return _switchtupleunion(𝕃, argtypes, na, [], nothing, groups)
... | nit: The implementation right now constrains that `i > j`, but nothing really says that it must be true>
```suggestion
for j in 1:length(t)
```
Additionally, when we have alias groups, it might be cool to similarly refine dependent types during `smerge!` also | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
end
end
function _hasmethod(@nospecialize(tt)) # this function has a special tfunc
world = ccall(:jl_get_tls_world_age, UInt, ()) # tls_world_age()
return Intrinsics.not_int(ccall(:jl_gf_invoke_lookup, Any, (Any, Any, UInt), tt, nothing, world) === nothing)
end
# for backward compat
arrayref(inbounds::Boo... | This is pretty easy to lead to mistakes in the future:
```julia
julia> Core.shl_int(0x1, 8)
0x00
``` | end
end
function _hasmethod(@nospecialize(tt)) # this function has a special tfunc
world = ccall(:jl_get_tls_world_age, UInt, ()) # tls_world_age()
return Intrinsics.not_int(ccall(:jl_gf_invoke_lookup, Any, (Any, Any, UInt), tt, nothing, world) === nothing)
end
# for backward compat
arrayref(inbounds::Boo... | @@ -1138,10 +1138,11 @@ EnterNode(old::EnterNode, new_dest::Int) = isdefined(old, :scope) ?
EnterNode(new_dest, old.scope) : EnterNode(new_dest)
# typename(_).constprop_heuristic
-const FORCE_CONST_PROP = 0x1
-const ARRAY_INDEX_HEURISTIC = 0x2
-const ITERATE_HEURISTIC = 0x3
-const SAMETYPE_HEURISTIC ... | base/boot.jl | 26 | Julia | 0.714 | suggestion | 100 | 51 | 51 | false | inference: make `constprop_heuristic` a bitfield | 61,045 | JuliaLang/julia | 35,573 | Julia | topolarity | aviatesk | Review this pull request change.
Language: Julia
File: base/boot.jl
PR Title: inference: make `constprop_heuristic` a bitfield
BEFORE CODE: end
end
function _hasmethod(@nospecialize(tt)) # this function has a special tfunc
world = ccall(:jl_get_tls_world_age, UInt, ()) # tls_world_age()
return Intrinsi... | This is pretty easy to lead to mistakes in the future:
```julia
julia> Core.shl_int(0x1, 8)
0x00
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
function force_const_prop(interp::AbstractInterpreter, @nospecialize(f), method::Method)
return is_aggressive_constprop(method) ||
InferenceParams(interp).aggressive_constant_propagation ||
!iszero(typename(typeof(f)).constprop_heuristic & Core.FORCE_CONST_PROP)
end
function const_prop_functi... | Doesn't the `elseif` here leave this as effectively exclusive with the above heuristic? | function force_const_prop(interp::AbstractInterpreter, @nospecialize(f), method::Method)
return is_aggressive_constprop(method) ||
InferenceParams(interp).aggressive_constant_propagation ||
!iszero(typename(typeof(f)).constprop_heuristic & Core.FORCE_CONST_PROP)
end
function const_prop_functi... | @@ -1179,14 +1185,14 @@ function const_prop_function_heuristic(interp::AbstractInterpreter, @nospecializ
elseif ⊑(𝕃ᵢ, arrty, Array) || ⊑(𝕃ᵢ, arrty, GenericMemory)
return false
end
- elseif heuristic === Core.ITERATE_HEURISTIC
+ elseif !iszero(heuristic & Core.I... | Compiler/src/abstractinterpretation.jl | 26 | Julia | 0.357 | question | 87 | 51 | 51 | false | inference: make `constprop_heuristic` a bitfield | 61,045 | JuliaLang/julia | 35,573 | Julia | topolarity | aviatesk | Review this pull request change.
Language: Julia
File: Compiler/src/abstractinterpretation.jl
PR Title: inference: make `constprop_heuristic` a bitfield
BEFORE CODE: function force_const_prop(interp::AbstractInterpreter, @nospecialize(f), method::Method)
return is_aggressive_constprop(method) ||
Infe... | Doesn't the `elseif` here leave this as effectively exclusive with the above heuristic? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
end
function lift_keyvalue_get!(compact::IncrementalCompact, idx::Int, stmt::Expr, 𝕃ₒ::AbstractLattice)
collection = stmt.args[end-1]
key = stmt.args[end]
leaves, visited_philikes = collect_leaves(compact, collection, Any, 𝕃ₒ, keyvalue_predecessors(key, 𝕃ₒ))
isempty(leaves) && return
lifted_le... | Is all of this extra checking useful?
Seems like the lifting is valid to do by the OptimizedGenerics contract regardless | end
function lift_keyvalue_get!(compact::IncrementalCompact, idx::Int, stmt::Expr, 𝕃ₒ::AbstractLattice)
collection = stmt.args[end-1]
key = stmt.args[end]
leaves, visited_philikes = collect_leaves(compact, collection, Any, 𝕃ₒ, keyvalue_predecessors(key, 𝕃ₒ))
isempty(leaves) && return
lifted_le... | @@ -1044,17 +1044,27 @@ function lift_keyvalue_get!(compact::IncrementalCompact, idx::Int, stmt::Expr,
result_t = tmerge(𝕃ₒ, result_t, argextype(v.val, compact))
end
+ # Extract the wrapper type (e.g. Some{V}) from the inferred return type
+ # Union{Nothing, Some{V}} by subtracting Nothing. Bail... | Compiler/src/ssair/passes.jl | 26 | Julia | 0.429 | performance | 121 | 51 | 51 | false | optimizer: Switch to `Some{}` in PersistentDict optimized generics | 61,053 | JuliaLang/julia | 35,573 | Julia | topolarity | Keno | Review this pull request change.
Language: Julia
File: Compiler/src/ssair/passes.jl
PR Title: optimizer: Switch to `Some{}` in PersistentDict optimized generics
BEFORE CODE: end
function lift_keyvalue_get!(compact::IncrementalCompact, idx::Int, stmt::Expr, 𝕃ₒ::AbstractLattice)
collection = stmt.args[end-1]
... | Is all of this extra checking useful?
Seems like the lifting is valid to do by the OptimizedGenerics contract regardless | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
source::LineNumberNode
end
struct Threw <: ExecutionResult
exception
current_exceptions::Base.ExceptionStack
source::LineNumberNode
end
function eval_test_comparison(comparison::Expr, quoted::Expr, source::LineNumberNode, negate::Bool=false)
comparison.head === :comparison || throw(ArgumentError("... | I think https://github.com/JuliaLang/julia/pull/61149 was right that `quoted` is the wrong predicate here in that it may have insufficiently many levels of quoting for correct printing. | source::LineNumberNode
end
struct Threw <: ExecutionResult
exception
current_exceptions::Base.ExceptionStack
source::LineNumberNode
end
function eval_test_comparison(comparison::Expr, ops::Vector{Any}, source::LineNumberNode, negate::Bool=false)
comparison.head === :comparison || throw(ArgumentErr... | @@ -368,33 +368,35 @@ struct Threw <: ExecutionResult
source::LineNumberNode
end
-function eval_test_comparison(comparison::Expr, quoted::Expr, source::LineNumberNode, negate::Bool=false)
+function eval_test_comparison(comparison::Expr, ops::Vector{Any}, source::LineNumberNode, negate::Bool=false)
comparis... | stdlib/Test/src/Test.jl | 26 | Julia | 0.5 | bug | 185 | 51 | 51 | false | Test: quote Symbol values in @test failure pretty printing | 61,054 | JuliaLang/julia | 35,573 | Julia | Keno | apoorvdarshan | Review this pull request change.
Language: Julia
File: stdlib/Test/src/Test.jl
PR Title: Test: quote Symbol values in @test failure pretty printing
BEFORE CODE: source::LineNumberNode
end
struct Threw <: ExecutionResult
exception
current_exceptions::Base.ExceptionStack
source::LineNumberNode
end
f... | I think https://github.com/JuliaLang/julia/pull/61149 was right that `quoted` is the wrong predicate here in that it may have insufficiently many levels of quoting for correct printing. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
if isnothing(st_k)
setattr!(newnode(graph, old_src, K"unknown_head", cs), :name_val, head_s)
else
newnode(graph, old_src, st_k, cs)
end
# elseif e isa GlobalRef
# TODO: Better-behaved as K"globalref", but lowering doesn't know this
else
# We may wa... | ```suggestion
function est_to_expr(st::SyntaxTree; suppress_linenodes=false)
```
just a suggestion, for clarity | if isnothing(st_k)
setattr!(newnode(graph, old_src, K"unknown_head", cs), :name_val, head_s)
else
newnode(graph, old_src, st_k, cs)
end
# elseif e isa GlobalRef
# TODO: Better-behaved as K"globalref", but lowering doesn't know this
else
# We may wa... | @@ -97,7 +97,7 @@ function _expr_to_est(graph::SyntaxGraph, @nospecialize(e), src::LineNumberNode)
return st._id, src
end
-function est_to_expr(st::SyntaxTree)
+function est_to_expr(st::SyntaxTree, suppress_linenodes=false) | JuliaLowering/src/compat.jl | 26 | Julia | 0.714 | suggestion | 111 | 51 | 51 | false | [JuliaLowering] Several small bugfixes found in stdlib testing | 61,070 | JuliaLang/julia | 35,573 | Julia | topolarity | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/compat.jl
PR Title: [JuliaLowering] Several small bugfixes found in stdlib testing
BEFORE CODE: if isnothing(st_k)
setattr!(newnode(graph, old_src, K"unknown_head", cs), :name_val, head_s)
else
newnode(g... | ```suggestion
function est_to_expr(st::SyntaxTree; suppress_linenodes=false)
```
just a suggestion, for clarity | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
QuoteNode(est_to_expr(st[1]))
else
# TODO: should handle post-lowering forms as well
@assert !is_leaf(st) "est_to_expr should only be used pre-desugaring"
# In a partially-expanded or quoted AST, there may be heads with no
# corresponding kind
head = Symbol(k === K"un... | ```suggestion
let suppress_linenodes = i == 1 && (k == K"for" || k == K"let")
push!(out.args, est_to_expr(c; suppress_linenodes))
end
``` | # Expr(st.value).
isa_lowering_ast_node(v) ? QuoteNode(v) : v
elseif k === K"inert"
QuoteNode(est_to_expr(st[1]))
else
# TODO: should handle post-lowering forms as well
@assert !is_leaf(st) "est_to_expr should only be used pre-desugaring"
# In a partially-expanded... | @@ -116,22 +116,45 @@ function est_to_expr(st::SyntaxTree)
elseif k === K"inert"
QuoteNode(est_to_expr(st[1]))
else
- # TODO: should handle post-desugaring forms as well
+ # TODO: should handle post-lowering forms as well
@assert !is_leaf(st) "est_to_expr should only be used pr... | JuliaLowering/src/compat.jl | 26 | Julia | 0.714 | suggestion | 177 | 51 | 51 | false | [JuliaLowering] Several small bugfixes found in stdlib testing | 61,070 | JuliaLang/julia | 35,573 | Julia | topolarity | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/compat.jl
PR Title: [JuliaLowering] Several small bugfixes found in stdlib testing
BEFORE CODE: QuoteNode(est_to_expr(st[1]))
else
# TODO: should handle post-lowering forms as well
@assert !is_leaf(st) "est_to_expr ... | ```suggestion
let suppress_linenodes = i == 1 && (k == K"for" || k == K"let")
push!(out.args, est_to_expr(c; suppress_linenodes))
end
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# Macro authors are responsible for handling any linenodes that follow
# the rules above (but the presence of optional linenodes can't be
# counted upon).
need_lnns = head in (:block, :toplevel) && !suppress_linenodes
for (i, c) in enumerate(children(st))
need_lnns &&... | Is it possible for this to add the `LineNumberNode` we're trying to suppress? | # Other than that, optional linenodes may show up anywhere within:
# - `block`, unless the block is the first child of `for` or `let`
# - `toplevel`
# Macro authors are responsible for handling any linenodes that follow
# the rules above (but the presence of optional linenodes ca... | @@ -116,22 +116,45 @@ function est_to_expr(st::SyntaxTree)
elseif k === K"inert"
QuoteNode(est_to_expr(st[1]))
else
- # TODO: should handle post-desugaring forms as well
+ # TODO: should handle post-lowering forms as well
@assert !is_leaf(st) "est_to_expr should only be used pr... | JuliaLowering/src/compat.jl | 26 | Julia | 0.357 | question | 77 | 51 | 51 | false | [JuliaLowering] Several small bugfixes found in stdlib testing | 61,070 | JuliaLang/julia | 35,573 | Julia | topolarity | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/compat.jl
PR Title: [JuliaLowering] Several small bugfixes found in stdlib testing
BEFORE CODE: # Macro authors are responsible for handling any linenodes that follow
# the rules above (but the presence of optional linenodes ca... | Is it possible for this to add the `LineNumberNode` we're trying to suppress? | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
head = Symbol(k === K"unknown_head" ? st.name_val : untokenize(k))
out = Expr(head)
# (Move the following assumptions to the docs if they turn out accurate)
# The only mandatory LineNumberNode is the second macrocall argument.
# Other than that, optional linenodes may show up an... | ```suggestion
# all following cases ignore `suppress_linenodes`
# (it only applies to the immediate expression, not transitively to children)
[K"module" _... [K"block" _...]] ->
``` | @assert !is_leaf(st) "est_to_expr should only be used pre-desugaring"
# In a partially-expanded or quoted AST, there may be heads with no
# corresponding kind
head = Symbol(k === K"unknown_head" ? st.name_val : untokenize(k))
out = Expr(head)
# (Move the following assump... | @@ -116,22 +116,45 @@ function est_to_expr(st::SyntaxTree)
elseif k === K"inert"
QuoteNode(est_to_expr(st[1]))
else
- # TODO: should handle post-desugaring forms as well
+ # TODO: should handle post-lowering forms as well
@assert !is_leaf(st) "est_to_expr should only be used pr... | JuliaLowering/src/compat.jl | 26 | Julia | 0.786 | suggestion | 217 | 51 | 51 | false | [JuliaLowering] Several small bugfixes found in stdlib testing | 61,070 | JuliaLang/julia | 35,573 | Julia | topolarity | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/compat.jl
PR Title: [JuliaLowering] Several small bugfixes found in stdlib testing
BEFORE CODE: head = Symbol(k === K"unknown_head" ? st.name_val : untokenize(k))
out = Expr(head)
# (Move the following assumptions to t... | ```suggestion
# all following cases ignore `suppress_linenodes`
# (it only applies to the immediate expression, not transitively to children)
[K"module" _... [K"block" _...]] ->
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
int sweep_full = next_sweep_full;
int recollect = 0;
assert(gc_n_threads != 0);
for (int i = 0; i < gc_n_threads; i++) {
jl_ptls_t ptls2 = gc_all_tls_states[i];
if (ptls2 != NULL)
remset_nptr += ptls2->gc_tls.heap.remset_nptr;
}
(void)remset_nptr; //Use this informati... | please take a note - we are monitoring content of `gc_record_full_sweep_reason` and gc_num during pre/post gc hooks to understand what runtime decided to do.
please make sure that we would be able to reconstruct that FULL GC scheduled but cancelled | int remset_nptr = 0;
int sweep_full = next_sweep_full;
int recollect = 0;
assert(gc_n_threads != 0);
for (int i = 0; i < gc_n_threads; i++) {
jl_ptls_t ptls2 = gc_all_tls_states[i];
if (ptls2 != NULL)
remset_nptr += ptls2->gc_tls.heap.remset_nptr;
}
(void)remset_n... | @@ -3177,6 +3178,9 @@ static int _jl_gc_collect(jl_ptls_t ptls, jl_gc_collection_t collection) JL_NOTS
recollect = 1;
gc_record_full_sweep_reason(FULL_SWEEP_REASON_FORCED_FULL_SWEEP);
}
+ if (gc_disable_auto_full_sweep && collection != JL_GC_FULL) {
+ sweep_full = 0; | src/gc-stock.c | 26 | C | 0.571 | suggestion | 248 | 51 | 51 | false | Add API to disable full collections | 61,087 | JuliaLang/julia | 35,573 | Julia | tz-lom | gbaraldi | Review this pull request change.
Language: C
File: src/gc-stock.c
PR Title: Add API to disable full collections
BEFORE CODE: int sweep_full = next_sweep_full;
int recollect = 0;
assert(gc_n_threads != 0);
for (int i = 0; i < gc_n_threads; i++) {
jl_ptls_t ptls2 = gc_all_tls_states[i];
... | please take a note - we are monitoring content of `gc_record_full_sweep_reason` and gc_num during pre/post gc hooks to understand what runtime decided to do.
please make sure that we would be able to reconstruct that FULL GC scheduled but cancelled | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
x == 2 && return Int(0)
x == 3 && return C_NULL
return nothing
end
function _bitcast_trigger(x::Int)
val = _bitcast_returns_union(x)
return Core.Intrinsics.bitcast(Ptr{Nothing}, val)
end
@test _bitcast_trigger(1) === Ptr{Nothing}(0)
@test _bitcast_trigger(3) =... | This triggers some UB and may make the GC segfault (since the unsafe_store isn't visible, so in rare cases the `val` might survive after the preserve block ends but not the contents). Can you use a `Libc.malloc` for buf to avoid any issues | x == 2 && return Int(0)
x == 3 && return C_NULL
return nothing
end
function _bitcast_trigger(x::Int)
val = _bitcast_returns_union(x)
return Core.Intrinsics.bitcast(Ptr{Nothing}, val)
end
@test _bitcast_trigger(1) === Ptr{Nothing}(0)
@test _bitcast_trigger(3) =... | @@ -587,3 +587,29 @@ end
@test _bitcast_trigger(1) === Ptr{Nothing}(0)
@test _bitcast_trigger(3) === Ptr{Nothing}(0)
end
+
+# Test unsafe_store! on union values with inline_roots (split representation)
+@testset "pointerset union with inline_roots" begin
+ struct PointersetMixedGC
+ a::Vector{Int}
... | test/intrinsics.jl | 26 | Julia | 0.643 | bug | 239 | 38 | 38 | false | fix pointerset for values with inline_roots | 61,105 | JuliaLang/julia | 35,573 | Julia | vtjnash | adienes | Review this pull request change.
Language: Julia
File: test/intrinsics.jl
PR Title: fix pointerset for values with inline_roots
BEFORE CODE: x == 2 && return Int(0)
x == 3 && return C_NULL
return nothing
end
function _bitcast_trigger(x::Int)
val = _bitcast_returns_union(x)
... | This triggers some UB and may make the GC segfault (since the unsafe_store isn't visible, so in rare cases the `val` might survive after the preserve block ends but not the contents). Can you use a `Libc.malloc` for buf to avoid any issues | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
for (id, from_outer_lambda) in lambda_bindings.locals_capt
b = get_binding(ctx, id)
!b.is_captured && continue
from_outer_lambda && continue
if b.is_assigned_once && b.kind in (:local, :argument)
push!(candidates, id)
end
end
isempty(candidates) && return
... | ```suggestion
# safe-to-unbox implies is-always-defined (but not vice-versa)
b.safe_to_unbox = true
b.is_always_defined = true
```
this may be clearer than the separate `for` loop below, since the is_always_defined check is not actually sufficient on its own, which is why we cannot ... | for (id, from_outer_lambda) in lambda_bindings.locals_capt
b = get_binding(ctx, id)
!b.is_captured && continue
from_outer_lambda && continue
if b.is_assigned_once && b.kind in (:local, :argument)
push!(candidates, id)
end
end
isempty(candidates) && return
... | @@ -319,40 +308,38 @@ function _analyze_lambda_vars!(ctx, ex)
# Collect candidate variables: captured and single-assigned
candidates = Set{IdTag}()
for (id, from_outer_lambda) in lambda_bindings.locals_capt
- binfo = get_binding(ctx, id)
- maybe_boxed = binfo.is_captured && binfo.kind in (:... | JuliaLowering/src/binding_analysis.jl | 26 | Julia | 0.857 | suggestion | 340 | 37 | 26 | false | [JuliaLowering] Don't clobber `is_always_defined` in unboxing | 61,134 | JuliaLang/julia | 35,573 | Julia | topolarity | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/binding_analysis.jl
PR Title: [JuliaLowering] Don't clobber `is_always_defined` in unboxing
BEFORE CODE: for (id, from_outer_lambda) in lambda_bindings.locals_capt
b = get_binding(ctx, id)
!b.is_captured && continue
... | ```suggestion
# safe-to-unbox implies is-always-defined (but not vice-versa)
b.safe_to_unbox = true
b.is_always_defined = true
```
this may be clearer than the separate `for` loop below, since the is_always_defined check is not actually sufficient on its own, which is why we cannot ... | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
symbolic_block_labels::Set{String} # labels that are symbolic blocks (not allowed as @goto targets)
meta::Dict{Symbol, Any}
mod::Module
end
function LinearIRContext(graph, ctx, is_toplevel_thunk, lambda_bindings, return_type)
rett = isnothing(return_type) ? nothing : reparent(graph, return_type)
A... | ```suggestion
binfo.is_always_defined
``` | symbolic_block_labels::Set{String} # labels that are symbolic blocks (not allowed as @goto targets)
meta::Dict{Symbol, Any}
mod::Module
end
function LinearIRContext(graph, ctx, is_toplevel_thunk, lambda_bindings, return_type)
rett = isnothing(return_type) ? nothing : reparent(graph, return_type)
A... | @@ -106,9 +106,6 @@ function is_valid_body_ir_argument(ctx, ex)
true
elseif kind(ex) == K"BindingId"
binfo = get_binding(ctx, ex)
- # arguments are inherently always-defined, but the closure
- # box analysis overrides this flag to mean "valid to leave
- # unboxed" so we check... | JuliaLowering/src/linear_ir.jl | 26 | Julia | 0.429 | suggestion | 49 | 51 | 51 | false | [JuliaLowering] Don't clobber `is_always_defined` in unboxing | 61,134 | JuliaLang/julia | 35,573 | Julia | topolarity | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/linear_ir.jl
PR Title: [JuliaLowering] Don't clobber `is_always_defined` in unboxing
BEFORE CODE: symbolic_block_labels::Set{String} # labels that are symbolic blocks (not allowed as @goto targets)
meta::Dict{Symbol, Any}
mod::Mod... | ```suggestion
binfo.is_always_defined
``` | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
# fields. The new type will be named `name_str` which must be an unassigned
# name in the module.
function type_for_closure(ctx::ClosureConversionCtx, srcref, name_str, field_syms, field_is_box)
# New closure types always belong to the module we're expanding into - they
# need to be serialized there during prec... | Arguably this should be set in the `.unboxed` analysis too | # fields. The new type will be named `name_str` which must be an unassigned
# name in the module.
function type_for_closure(ctx::ClosureConversionCtx, srcref, name_str, field_syms, field_is_box)
# New closure types always belong to the module we're expanding into - they
# need to be serialized there during prec... | @@ -292,20 +292,15 @@ function type_for_closure(ctx::ClosureConversionCtx, srcref, name_str, field_sym
type_ex, type_binding
end
+# No box needed for:
+# - non-captured vars
+# - static params (can't be reassigned)
+# - any local our optimizations have determined to be unboxed
function is_boxed(binfo::BindingI... | JuliaLowering/src/closure_conversion.jl | 26 | Julia | 0.429 | suggestion | 58 | 51 | 51 | false | [JuliaLowering] Don't clobber `is_always_defined` in unboxing | 61,134 | JuliaLang/julia | 35,573 | Julia | topolarity | mlechu | Review this pull request change.
Language: Julia
File: JuliaLowering/src/closure_conversion.jl
PR Title: [JuliaLowering] Don't clobber `is_always_defined` in unboxing
BEFORE CODE: # fields. The new type will be named `name_str` which must be an unassigned
# name in the module.
function type_for_closure(ctx::Closure... | Arguably this should be set in the `.unboxed` analysis too | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
}
jl_locked_stream &get_dump_compiles_stream() JL_NOTSAFEPOINT JL_NOTSAFEPOINT_ENTER {
return dump_compiles_stream;
}
jl_locked_stream &get_dump_llvm_opt_stream() JL_NOTSAFEPOINT JL_NOTSAFEPOINT_ENTER {
return dump_llvm_opt_stream;
}
std::string getMangledName(StringRef Name) JL_... | Would be good to document the meaning of the return value here. | }
jl_locked_stream &get_dump_compiles_stream() JL_NOTSAFEPOINT JL_NOTSAFEPOINT_ENTER {
return dump_compiles_stream;
}
jl_locked_stream &get_dump_llvm_opt_stream() JL_NOTSAFEPOINT JL_NOTSAFEPOINT_ENTER {
return dump_llvm_opt_stream;
}
std::string getMangledName(StringRef Name) JL_... | @@ -785,7 +785,7 @@ class JuliaOJIT {
// Rename LinkGraph symbols to match the previously chosen names and
// register debug info for defined symbols.
- void linkOutput(orc::MaterializationResponsibility &MR, MemoryBufferRef ObjBuf,
+ bool linkOutput(orc::MaterializationResponsibility &MR, MemoryBuffe... | src/jitlayers.h | 26 | C/C++ | 0.214 | suggestion | 63 | 51 | 51 | false | Local names linking fixes | 61,165 | JuliaLang/julia | 35,573 | Julia | topolarity | xal-0 | Review this pull request change.
Language: C/C++
File: src/jitlayers.h
PR Title: Local names linking fixes
BEFORE CODE: }
jl_locked_stream &get_dump_compiles_stream() JL_NOTSAFEPOINT JL_NOTSAFEPOINT_ENTER {
return dump_compiles_stream;
}
jl_locked_stream &get_dump_llvm_opt_stream() JL_NOTSAF... | Would be good to document the meaning of the return value here. | You are an expert software engineer and code reviewer. Review pull requests carefully for correctness, bugs, security issues, performance problems, maintainability, and unnecessary complexity. Only raise actionable issues supported by the code. Explain why an issue matters and suggest a concrete fix when appropriate. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.