repo
stringlengths
5
53
pr_number
int32
1
321k
task_type
stringclasses
2 values
issue_text
stringlengths
0
81.2k
pr_title
stringlengths
1
319
pr_body
stringlengths
0
105k
base_sha
stringlengths
40
40
head_sha
stringlengths
40
40
gold_diff
stringlengths
0
202M
changed_files
listlengths
0
100
review_threads
listlengths
0
100
test_patch
stringlengths
0
23.4M
merged
bool
1 class
wasmerio/wasmer
6,658
issue_to_patch
feat(V8): enable some WASIX tests on Windows
The PR enables testing V8 for the compiler tests and some of the WASIX tests (note `wasixcc` does not support Windows yet). A proper enabling of the V8 for wasmer-cli will come in a subsequent PR. Issue: #6121
822170748615cc1e378b0063c6e70dc9091056e4
53dc199bd69b2dc378a856952b691cb64a911f61
diff --git a/Makefile b/Makefile index 187f70dc8a5b..1fad24bd5858 100644 --- a/Makefile +++ b/Makefile @@ -220,6 +220,10 @@ endif compilers := $(strip $(compilers)) build_compilers := $(strip $(build_compilers) $(compilers)) +ifeq ($(IS_WINDOWS), 1) + build_compilers := $(filter-out llvm,$(build_compilers)) +endif ...
[ "Makefile", "lib/api/build.rs", "lib/wasix/Cargo.toml", "lib/wasix/tests/wasm_tests/mod.rs", "lib/wasix/tests/wasm_tests/runner.rs", "lib/wasix/tests/wasm_tests/wasi_fyi/fs_sandbox_symlink.rs", "lib/wasix/tests/wasm_tests/wasi_fyi/ported_readlink.rs", "tests/lib/wast/src/wast.rs" ]
[ { "comment": "Removing the Windows short-circuit makes these fixture-based WASIX integration tests run on Windows, but the harness currently relies on Unix tooling (e.g., `bash` in `run_build_script`) and will also error on symlink fixtures (`copy_symlink` bails on non-Unix). This will cause `cargo test`/`make ...
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index dea874f7b8b3..db1d9db8e70b 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -119,6 +119,7 @@ struct MappedDirectory { #[derive(Debug, Clone, Copy, PartialEq, Eq)] pub enum Engine { Cr...
true
wasmerio/wasmer
6,658
comment_to_fix
feat(V8): enable some WASIX tests on Windows
Removing the Windows short-circuit makes these fixture-based WASIX integration tests run on Windows, but the harness currently relies on Unix tooling (e.g., `bash` in `run_build_script`) and will also error on symlink fixtures (`copy_symlink` bails on non-Unix). This will cause `cargo test`/`make test-all` to fail on W...
822170748615cc1e378b0063c6e70dc9091056e4
53dc199bd69b2dc378a856952b691cb64a911f61
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index dea874f7b8b3..db1d9db8e70b 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -119,6 +119,7 @@ struct MappedDirectory { #[derive(Debug, Clone, Copy, PartialEq, Eq)] pub enum Engine { Cr...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "Removing the Windows short-circuit makes these fixture-based WASIX integration tests run on Windows, but the harness currently relies on Unix tooling (e.g., `bash` in `run_build_script`) and will also error on symlink fixtures (`copy_symlink` bails on non-Unix). This will cause `cargo test`/`make ...
true
wasmerio/wasmer
6,660
issue_to_patch
fix(cargo): publish
b14f2e731c0c9c426496b0703caf0a2933e2eba4
5a8150144438b2ada347b6d02a5ea658e89b8bc6
diff --git a/lib/swift/Cargo.toml b/lib/swift/Cargo.toml index f41cee14d500..c8da5cc93c1c 100644 --- a/lib/swift/Cargo.toml +++ b/lib/swift/Cargo.toml @@ -19,6 +19,7 @@ virtual-fs = { path = "../virtual-fs", version = "=0.702.0-alpha.3", default-fea "webc-fs", ] } wasmer = { version = "=7.2.0-alpha.3", path = "../...
[ "lib/swift/Cargo.toml", "lib/wasix/Cargo.toml" ]
[ { "comment": "`exclude = [\"tests/wasm_tests/**\"]` will also exclude the `tests/wasm_tests/mod.rs` file referenced by the explicit `[[test]]` target below. That makes the packaged crate inconsistent (and can cause `cargo test --features sys-thread` to fail for consumers, or even `cargo package`/`cargo publish`...
true
wasmerio/wasmer
6,660
comment_to_fix
fix(cargo): publish
`exclude = ["tests/wasm_tests/**"]` will also exclude the `tests/wasm_tests/mod.rs` file referenced by the explicit `[[test]]` target below. That makes the packaged crate inconsistent (and can cause `cargo test --features sys-thread` to fail for consumers, or even `cargo package`/`cargo publish` verification to error d...
b14f2e731c0c9c426496b0703caf0a2933e2eba4
5a8150144438b2ada347b6d02a5ea658e89b8bc6
diff --git a/lib/wasix/Cargo.toml b/lib/wasix/Cargo.toml index c7f7a156bd11..173cc4ccda58 100644 --- a/lib/wasix/Cargo.toml +++ b/lib/wasix/Cargo.toml @@ -11,6 +11,7 @@ homepage.workspace = true license.workspace = true repository.workspace = true rust-version.workspace = true +exclude = ["tests/wasm_tests/*/**"] ...
[ "lib/wasix/Cargo.toml" ]
[ { "comment": "`exclude = [\"tests/wasm_tests/**\"]` will also exclude the `tests/wasm_tests/mod.rs` file referenced by the explicit `[[test]]` target below. That makes the packaged crate inconsistent (and can cause `cargo test --features sys-thread` to fail for consumers, or even `cargo package`/`cargo publish`...
true
wasmerio/wasmer
6,640
issue_to_patch
perf(vm): thread-local cache for on_wasm_stack coroutine stacks
The hot path in on_wasm_stack pops and pushes coroutine stacks through a global crossbeam_queue::SegQueue. On a tight loop of typed Wasm calls, the atomic CAS on every pop/push dominates the call: SegQueue::pop ~23% and SegQueue::push ~17% of total cycles measured via perf record on a single-export bench. Add a thre...
b14f2e731c0c9c426496b0703caf0a2933e2eba4
76a3ce2751c2d60577f93aa302bf68a740343816
diff --git a/Cargo.toml b/Cargo.toml index 160f7e61b8f3..3dce0c299e26 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -446,6 +446,10 @@ harness = false name = "import_functions" harness = false +[[bench]] +name = "typed_call" +harness = false + [[example]] name = "early-exit" path = "examples/early_exit.rs" diff --g...
[ "Cargo.toml", "benches/typed_call.rs", "lib/vm/src/trap/traphandlers.rs" ]
[ { "comment": "In a re-entrant/nested `on_wasm_stack` call on the same thread, the outer call has already taken the TLS slot (now `None`), so the inner call's `release_stack` will place its stack into the TLS slot. When the outer call then completes and tries to `release_stack`, it will displace the inner's stac...
true
wasmerio/wasmer
6,640
comment_to_fix
perf(vm): thread-local cache for on_wasm_stack coroutine stacks
In a re-entrant/nested `on_wasm_stack` call on the same thread, the outer call has already taken the TLS slot (now `None`), so the inner call's `release_stack` will place its stack into the TLS slot. When the outer call then completes and tries to `release_stack`, it will displace the inner's stack to the global `STACK...
b14f2e731c0c9c426496b0703caf0a2933e2eba4
76a3ce2751c2d60577f93aa302bf68a740343816
diff --git a/lib/vm/src/trap/traphandlers.rs b/lib/vm/src/trap/traphandlers.rs index 015c8b6c8d6e..8022c52211a9 100644 --- a/lib/vm/src/trap/traphandlers.rs +++ b/lib/vm/src/trap/traphandlers.rs @@ -88,22 +88,85 @@ pub fn get_stack_size() -> usize { } /// Pool of pre-allocated coroutine stacks to avoid repeated mma...
[ "lib/vm/src/trap/traphandlers.rs" ]
[ { "comment": "In a re-entrant/nested `on_wasm_stack` call on the same thread, the outer call has already taken the TLS slot (now `None`), so the inner call's `release_stack` will place its stack into the TLS slot. When the outer call then completes and tries to `release_stack`, it will displace the inner's stac...
true
wasmerio/wasmer
6,640
comment_to_fix
perf(vm): thread-local cache for on_wasm_stack coroutine stacks
This test (and `tls_stack_caches_after_first_call`) leaves no explicit cleanup of `TLS_STACK` on the test-runner thread for the first test, and the cleanup in `tls_stack_caches_after_first_call` (line 1490) only clears the slot — the stack is dropped rather than returned to the pool. Since tests share threads in the de...
b14f2e731c0c9c426496b0703caf0a2933e2eba4
76a3ce2751c2d60577f93aa302bf68a740343816
diff --git a/lib/vm/src/trap/traphandlers.rs b/lib/vm/src/trap/traphandlers.rs index 015c8b6c8d6e..8022c52211a9 100644 --- a/lib/vm/src/trap/traphandlers.rs +++ b/lib/vm/src/trap/traphandlers.rs @@ -88,22 +88,85 @@ pub fn get_stack_size() -> usize { } /// Pool of pre-allocated coroutine stacks to avoid repeated mma...
[ "lib/vm/src/trap/traphandlers.rs" ]
[ { "comment": "This test (and `tls_stack_caches_after_first_call`) leaves no explicit cleanup of `TLS_STACK` on the test-runner thread for the first test, and the cleanup in `tls_stack_caches_after_first_call` (line 1490) only clears the slot — the stack is dropped rather than returned to the pool. Since tests s...
true
wasmerio/wasmer
6,640
comment_to_fix
perf(vm): thread-local cache for on_wasm_stack coroutine stacks
This test holds `GLOBAL_STATE` lock on the main thread while the spawned thread runs `on_wasm_stack`. If `on_wasm_stack` (or any code it transitively calls) ever tries to acquire `GLOBAL_STATE`, this will deadlock. It's fine today, but consider releasing the lock before `handle.join()` or documenting why the spawned th...
b14f2e731c0c9c426496b0703caf0a2933e2eba4
76a3ce2751c2d60577f93aa302bf68a740343816
diff --git a/lib/vm/src/trap/traphandlers.rs b/lib/vm/src/trap/traphandlers.rs index 015c8b6c8d6e..8022c52211a9 100644 --- a/lib/vm/src/trap/traphandlers.rs +++ b/lib/vm/src/trap/traphandlers.rs @@ -88,22 +88,85 @@ pub fn get_stack_size() -> usize { } /// Pool of pre-allocated coroutine stacks to avoid repeated mma...
[ "lib/vm/src/trap/traphandlers.rs" ]
[ { "comment": "This test holds `GLOBAL_STATE` lock on the main thread while the spawned thread runs `on_wasm_stack`. If `on_wasm_stack` (or any code it transitively calls) ever tries to acquire `GLOBAL_STATE`, this will deadlock. It's fine today, but consider releasing the lock before `handle.join()` or document...
true
wasmerio/wasmer
6,645
issue_to_patch
tests: run WASI FYI with Singlepass as well
Issue: #6566
caf909f12ce8124141fb1d32c12c51ad8b335e20
bc88c5b739b1e248f932478a656af484454f8d0d
diff --git a/lib/wasix/Cargo.toml b/lib/wasix/Cargo.toml index 170d5b8c05f8..8c28641d1ac7 100644 --- a/lib/wasix/Cargo.toml +++ b/lib/wasix/Cargo.toml @@ -233,6 +233,7 @@ journal = ["tokio/fs", "wasmer-journal/log-file"] # Deprecated. Kept it for compatibility compiler = [] +singlepass = ["wasmer/singlepass"] v8 =...
[ "lib/wasix/Cargo.toml", "lib/wasix/tests/wasm_tests/mod.rs", "lib/wasix/tests/wasm_tests/runner.rs" ]
[ { "comment": "`OsStr` is imported unconditionally but only used inside `#[cfg(feature = \"singlepass\")]` blocks. When the `singlepass` feature is disabled this becomes an unused import and will fail under the repo's `-D unused-imports`/clippy settings.", "path": "lib/wasix/tests/wasm_tests/mod.rs", "hu...
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index d64104af823c..d0ea4b788c9e 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -40,7 +40,7 @@ //! `Ignored:{reason}` marks the configuration as ignored with the given reason. //! //! `SkipEn...
true
wasmerio/wasmer
6,645
comment_to_fix
tests: run WASI FYI with Singlepass as well
`OsStr` is imported unconditionally but only used inside `#[cfg(feature = "singlepass")]` blocks. When the `singlepass` feature is disabled this becomes an unused import and will fail under the repo's `-D unused-imports`/clippy settings.
caf909f12ce8124141fb1d32c12c51ad8b335e20
bc88c5b739b1e248f932478a656af484454f8d0d
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index d64104af823c..d0ea4b788c9e 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -40,7 +40,7 @@ //! `Ignored:{reason}` marks the configuration as ignored with the given reason. //! //! `SkipEn...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "`OsStr` is imported unconditionally but only used inside `#[cfg(feature = \"singlepass\")]` blocks. When the `singlepass` feature is disabled this becomes an unused import and will fail under the repo's `-D unused-imports`/clippy settings.", "path": "lib/wasix/tests/wasm_tests/mod.rs", "hu...
true
wasmerio/wasmer
6,645
comment_to_fix
tests: run WASI FYI with Singlepass as well
`create_engine_for_wasm` explicitly limits compilation to a single thread to avoid contention when many tests run in parallel, but the new Singlepass arm doesn't apply the same `num_threads(1)` setting. This can increase CPU usage / reduce test stability compared to LLVM/Cranelift.
caf909f12ce8124141fb1d32c12c51ad8b335e20
bc88c5b739b1e248f932478a656af484454f8d0d
diff --git a/lib/wasix/tests/wasm_tests/runner.rs b/lib/wasix/tests/wasm_tests/runner.rs index 73bd1f77a282..60789e29f133 100644 --- a/lib/wasix/tests/wasm_tests/runner.rs +++ b/lib/wasix/tests/wasm_tests/runner.rs @@ -265,6 +265,8 @@ fn create_engine_for_wasm(wasm_bytes: &[u8], engine: Engine) -> wasmer::Engine { ...
[ "lib/wasix/tests/wasm_tests/runner.rs" ]
[ { "comment": "`create_engine_for_wasm` explicitly limits compilation to a single thread to avoid contention when many tests run in parallel, but the new Singlepass arm doesn't apply the same `num_threads(1)` setting. This can increase CPU usage / reduce test stability compared to LLVM/Cranelift.", "path": "...
true
wasmerio/wasmer
6,645
comment_to_fix
tests: run WASI FYI with Singlepass as well
Now that `Engine::Singlepass` is a runnable engine, the `SkipEngine:{engine}:{reason}` directive should be able to recognize `singlepass` too. Currently the directive parser only accepts `llvm`, `cranelift`, and `v8`, so fixtures can’t selectively skip failing configs for Singlepass without ignoring them for all engine...
caf909f12ce8124141fb1d32c12c51ad8b335e20
bc88c5b739b1e248f932478a656af484454f8d0d
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index d64104af823c..d0ea4b788c9e 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -40,7 +40,7 @@ //! `Ignored:{reason}` marks the configuration as ignored with the given reason. //! //! `SkipEn...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "Now that `Engine::Singlepass` is a runnable engine, the `SkipEngine:{engine}:{reason}` directive should be able to recognize `singlepass` too. Currently the directive parser only accepts `llvm`, `cranelift`, and `v8`, so fixtures can’t selectively skip failing configs for Singlepass without ignori...
true
wasmerio/wasmer
6,467
issue_to_patch
WASIX: fix writing to unlinked files with reference counting and POSIX-compliant unlink()/rmdir() API We used to have a PR that only served as a TODO list placeholder: #6129
Fix FDs becoming invalid once their file is unlinked
Fixes: #6360.
cd2344ac409e2981dda47b534b5d0ce59deae876
2bf777c8cc2c6abde45eb1ced04bc67fc71cfa27
diff --git a/lib/virtual-fs/src/lib.rs b/lib/virtual-fs/src/lib.rs index 7fbc71182649..0f45c3549b0e 100644 --- a/lib/virtual-fs/src/lib.rs +++ b/lib/virtual-fs/src/lib.rs @@ -361,7 +361,11 @@ pub trait VirtualFile: /// the extra bytes will be allocated and zeroed fn set_len(&mut self, new_size: u64) -> Result...
[ "lib/virtual-fs/src/lib.rs", "lib/virtual-fs/src/mem_fs/file.rs", "lib/virtual-fs/src/mem_fs/file_opener.rs", "lib/virtual-fs/src/mem_fs/filesystem.rs", "lib/virtual-fs/src/mem_fs/mod.rs", "lib/virtual-fs/src/overlay_fs.rs", "tests/wasix/unlink-open-fd-write-after-unlink/main.c", "tests/wasix/unlink-o...
[ { "comment": "Open-handle tracking is used to decide whether `unlink_file_inode` can remove the inode immediately, but `lifecycle.opened()` is called inside `FileHandle::new()` after the filesystem write-lock has been released. This leaves a race where another thread can unlink the file between dropping the loc...
diff --git a/tests/wasix/unlink-open-fd-write-after-unlink/main.c b/tests/wasix/unlink-open-fd-write-after-unlink/main.c new file mode 100644 index 000000000000..90c8fd855de6 --- /dev/null +++ b/tests/wasix/unlink-open-fd-write-after-unlink/main.c @@ -0,0 +1,165 @@ +#include <errno.h> +#include <fcntl.h> +#include <std...
true
wasmerio/wasmer
6,467
comment_to_fix
Fix FDs becoming invalid once their file is unlinked
Open-handle tracking is used to decide whether `unlink_file_inode` can remove the inode immediately, but `lifecycle.opened()` is called inside `FileHandle::new()` after the filesystem write-lock has been released. This leaves a race where another thread can unlink the file between dropping the lock and incrementing `op...
cd2344ac409e2981dda47b534b5d0ce59deae876
2bf777c8cc2c6abde45eb1ced04bc67fc71cfa27
diff --git a/lib/virtual-fs/src/mem_fs/file_opener.rs b/lib/virtual-fs/src/mem_fs/file_opener.rs index fac028187519..2047b00e6e00 100644 --- a/lib/virtual-fs/src/mem_fs/file_opener.rs +++ b/lib/virtual-fs/src/mem_fs/file_opener.rs @@ -51,6 +51,7 @@ impl FileSystem { len: file_len, ...
[ "lib/virtual-fs/src/mem_fs/file_opener.rs" ]
[ { "comment": "Open-handle tracking is used to decide whether `unlink_file_inode` can remove the inode immediately, but `lifecycle.opened()` is called inside `FileHandle::new()` after the filesystem write-lock has been released. This leaves a race where another thread can unlink the file between dropping the loc...
true
wasmerio/wasmer
6,467
comment_to_fix
Fix FDs becoming invalid once their file is unlinked
`handle_lifecycle` extraction treats `InodeResolution::Redirect(..)` as `unreachable!()`, but `maybe_inode_of_file` can legitimately be a redirect (e.g. when the parent is an `ArcDirectory`). This will panic on open rather than following the redirect logic handled later in the function. Handle the redirect here (e.g. e...
cd2344ac409e2981dda47b534b5d0ce59deae876
2bf777c8cc2c6abde45eb1ced04bc67fc71cfa27
diff --git a/lib/virtual-fs/src/mem_fs/file_opener.rs b/lib/virtual-fs/src/mem_fs/file_opener.rs index fac028187519..2047b00e6e00 100644 --- a/lib/virtual-fs/src/mem_fs/file_opener.rs +++ b/lib/virtual-fs/src/mem_fs/file_opener.rs @@ -51,6 +51,7 @@ impl FileSystem { len: file_len, ...
[ "lib/virtual-fs/src/mem_fs/file_opener.rs" ]
[ { "comment": "`handle_lifecycle` extraction treats `InodeResolution::Redirect(..)` as `unreachable!()`, but `maybe_inode_of_file` can legitimately be a redirect (e.g. when the parent is an `ArcDirectory`). This will panic on open rather than following the redirect logic handled later in the function. Handle the...
true
wasmerio/wasmer
6,467
comment_to_fix
Fix FDs becoming invalid once their file is unlinked
`SecondaryFile` wraps the underlying `VirtualFile`, but this `VirtualFile` impl only forwards a subset of methods. Because `VirtualFile` provides defaults for methods like `get_special_fd()`, `is_open()`, `as_owned_buffer()`, and `write_from_mmap()`, the wrapper will silently change behavior for secondary files that ov...
cd2344ac409e2981dda47b534b5d0ce59deae876
2bf777c8cc2c6abde45eb1ced04bc67fc71cfa27
diff --git a/lib/virtual-fs/src/overlay_fs.rs b/lib/virtual-fs/src/overlay_fs.rs index 0eff1b11d430..de9366bf4785 100644 --- a/lib/virtual-fs/src/overlay_fs.rs +++ b/lib/virtual-fs/src/overlay_fs.rs @@ -17,6 +17,168 @@ use crate::{ ReadDir, VirtualFile, ops, }; +fn unlink_overlay_path<P>(primary: &Arc<P>, path:...
[ "lib/virtual-fs/src/overlay_fs.rs" ]
[ { "comment": "`SecondaryFile` wraps the underlying `VirtualFile`, but this `VirtualFile` impl only forwards a subset of methods. Because `VirtualFile` provides defaults for methods like `get_special_fd()`, `is_open()`, `as_owned_buffer()`, and `write_from_mmap()`, the wrapper will silently change behavior for s...
true
wasmerio/wasmer
6,467
comment_to_fix
Fix FDs becoming invalid once their file is unlinked
`FileLifecycle::closed()` uses `fetch_sub(1) - 1`, which will underflow (wrapping to `usize::MAX`) if `closed()` is ever called when the count is 0 (e.g., if a handle is dropped without a matching `opened()` due to an early-return path or future refactor). This can mask accounting bugs and prevent storage reclamation. ...
cd2344ac409e2981dda47b534b5d0ce59deae876
2bf777c8cc2c6abde45eb1ced04bc67fc71cfa27
diff --git a/lib/virtual-fs/src/mem_fs/mod.rs b/lib/virtual-fs/src/mem_fs/mod.rs index db2ba68b625e..a1363737bb8d 100644 --- a/lib/virtual-fs/src/mem_fs/mod.rs +++ b/lib/virtual-fs/src/mem_fs/mod.rs @@ -17,7 +17,10 @@ use crate::Metadata; use std::{ ffi::{OsStr, OsString}, path::PathBuf, - sync::{Arc, Mut...
[ "lib/virtual-fs/src/mem_fs/mod.rs" ]
[ { "comment": "`FileLifecycle::closed()` uses `fetch_sub(1) - 1`, which will underflow (wrapping to `usize::MAX`) if `closed()` is ever called when the count is 0 (e.g., if a handle is dropped without a matching `opened()` due to an early-return path or future refactor). This can mask accounting bugs and prevent...
true
wasmerio/wasmer
6,467
comment_to_fix
Fix FDs becoming invalid once their file is unlinked
The `OnceLock` + global `Mutex<Option<...>>` hook introduces shared mutable global state across the entire test binary. Since Rust tests run in parallel by default, this can make tests flaky (a different test could trigger `run_open_before_handle_hook()` while the hook is set, or a panic could leave a hook installed fo...
cd2344ac409e2981dda47b534b5d0ce59deae876
2bf777c8cc2c6abde45eb1ced04bc67fc71cfa27
diff --git a/lib/virtual-fs/src/mem_fs/file_opener.rs b/lib/virtual-fs/src/mem_fs/file_opener.rs index fac028187519..2047b00e6e00 100644 --- a/lib/virtual-fs/src/mem_fs/file_opener.rs +++ b/lib/virtual-fs/src/mem_fs/file_opener.rs @@ -51,6 +51,7 @@ impl FileSystem { len: file_len, ...
[ "lib/virtual-fs/src/mem_fs/file_opener.rs" ]
[ { "comment": "The `OnceLock` + global `Mutex<Option<...>>` hook introduces shared mutable global state across the entire test binary. Since Rust tests run in parallel by default, this can make tests flaky (a different test could trigger `run_open_before_handle_hook()` while the hook is set, or a panic could lea...
true
wasmerio/wasmer
6,467
comment_to_fix
Fix FDs becoming invalid once their file is unlinked
`unlink_overlay_path` discards the specific error from `ops::create_white_out()` and then may return `FsError::PermissionDenied` when `primary.remove_file()` fails with `EntryNotFound`/`InvalidInput`/`BaseNotDirectory`. This can produce misleading results, e.g. if the whiteout already exists (`create_white_out` returns...
cd2344ac409e2981dda47b534b5d0ce59deae876
2bf777c8cc2c6abde45eb1ced04bc67fc71cfa27
diff --git a/lib/virtual-fs/src/overlay_fs.rs b/lib/virtual-fs/src/overlay_fs.rs index 0eff1b11d430..de9366bf4785 100644 --- a/lib/virtual-fs/src/overlay_fs.rs +++ b/lib/virtual-fs/src/overlay_fs.rs @@ -17,6 +17,168 @@ use crate::{ ReadDir, VirtualFile, ops, }; +fn unlink_overlay_path<P>(primary: &Arc<P>, path:...
[ "lib/virtual-fs/src/overlay_fs.rs" ]
[ { "comment": "`unlink_overlay_path` discards the specific error from `ops::create_white_out()` and then may return `FsError::PermissionDenied` when `primary.remove_file()` fails with `EntryNotFound`/`InvalidInput`/`BaseNotDirectory`. This can produce misleading results, e.g. if the whiteout already exists (`cre...
true
wasmerio/wasmer
6,467
comment_to_fix
Fix FDs becoming invalid once their file is unlinked
This helper only has the pathname, so an FD opened from a secondary/COW file will unlink whatever is currently bound to that path. If the path gets recreated before the old FD calls `unlink()`, the old handle will delete or whiteout the replacement instead of detaching the originally opened object. That is a real corre...
cd2344ac409e2981dda47b534b5d0ce59deae876
2bf777c8cc2c6abde45eb1ced04bc67fc71cfa27
diff --git a/lib/virtual-fs/src/overlay_fs.rs b/lib/virtual-fs/src/overlay_fs.rs index 0eff1b11d430..de9366bf4785 100644 --- a/lib/virtual-fs/src/overlay_fs.rs +++ b/lib/virtual-fs/src/overlay_fs.rs @@ -17,6 +17,168 @@ use crate::{ ReadDir, VirtualFile, ops, }; +fn unlink_overlay_path<P>(primary: &Arc<P>, path:...
[ "lib/virtual-fs/src/overlay_fs.rs" ]
[ { "comment": "This helper only has the pathname, so an FD opened from a secondary/COW file will unlink whatever is currently bound to that path. If the path gets recreated before the old FD calls `unlink()`, the old handle will delete or whiteout the replacement instead of detaching the originally opened object...
true
wasmerio/wasmer
6,467
comment_to_fix
Fix FDs becoming invalid once their file is unlinked
If `create_white_out()` failed but `primary.remove_file(path)` succeeds, this still returns success via the `other => other` arm. In that case the lower-layer file becomes visible again, so the overlay unlink did not actually succeed. Whiteout creation needs to be treated as part of the operation whenever a lower-layer...
cd2344ac409e2981dda47b534b5d0ce59deae876
2bf777c8cc2c6abde45eb1ced04bc67fc71cfa27
diff --git a/lib/virtual-fs/src/overlay_fs.rs b/lib/virtual-fs/src/overlay_fs.rs index 0eff1b11d430..de9366bf4785 100644 --- a/lib/virtual-fs/src/overlay_fs.rs +++ b/lib/virtual-fs/src/overlay_fs.rs @@ -17,6 +17,168 @@ use crate::{ ReadDir, VirtualFile, ops, }; +fn unlink_overlay_path<P>(primary: &Arc<P>, path:...
[ "lib/virtual-fs/src/overlay_fs.rs" ]
[ { "comment": "If `create_white_out()` failed but `primary.remove_file(path)` succeeds, this still returns success via the `other => other` arm. In that case the lower-layer file becomes visible again, so the overlay unlink did not actually succeed. Whiteout creation needs to be treated as part of the operation ...
true
wasmerio/wasmer
6,659
issue_to_patch
Release 7.2.0-alpha.3
[bot] Release wasmer version 7.2.0-alpha.3
cd2344ac409e2981dda47b534b5d0ce59deae876
cd08dcc5784cfa71aad3717d4f560c5d83cb4f59
diff --git a/CHANGELOG.md b/CHANGELOG.md index 37899ec72db7..5c64001b9efd 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -6,6 +6,92 @@ ## **Unreleased** +## 7.2.0-alpha.3 - 01/06/2026 + +## Added + + - [#6604](https://github.com/wasmerio/wasmer/pull/6604) Add soft-float libcall support for LLVM backend on hardwa...
[ "CHANGELOG.md", "Cargo.lock", "Cargo.toml", "lib/api/Cargo.toml", "lib/backend-api/Cargo.toml", "lib/c-api/Cargo.toml", "lib/cache/Cargo.toml", "lib/cli/Cargo.toml", "lib/compiler-cranelift/Cargo.toml", "lib/compiler-llvm/Cargo.toml", "lib/compiler-singlepass/Cargo.toml", "lib/compiler/Cargo.t...
[]
diff --git a/tests/integration/cli/Cargo.toml b/tests/integration/cli/Cargo.toml index a98b675655a3..5e2262ce101b 100644 --- a/tests/integration/cli/Cargo.toml +++ b/tests/integration/cli/Cargo.toml @@ -1,6 +1,6 @@ [package] name = "wasmer-integration-tests-cli" -version = "7.2.0-alpha.2" +version = "7.2.0-alpha.3" ...
true
wasmerio/wasmer
6,652
issue_to_patch
fix(virtual-fs): correct mem_fs poll_write_ready behavior
# Description `poll_write_ready()` in `mem_fs` was checking read perms, so write-only handles came back `PermissionDenied`. `CustomFile` and `ArcFile` also got forwarded to `poll_read_ready()`, so write readiness was kinda wrong there too. Repro before this patch: `cargo test -p virtual-fs poll_write_ready -- --nocapt...
a21e070815ac2cb24ed9475c3f35d03445c0344a
1c68a734d53d8f7e9b10621c04596b642553ce09
diff --git a/lib/virtual-fs/src/mem_fs/file.rs b/lib/virtual-fs/src/mem_fs/file.rs index 91767c6c5b87..9e1b4b00f806 100644 --- a/lib/virtual-fs/src/mem_fs/file.rs +++ b/lib/virtual-fs/src/mem_fs/file.rs @@ -441,11 +441,11 @@ impl VirtualFile for FileHandle { } fn poll_write_ready(mut self: Pin<&mut Self>, c...
[ "lib/virtual-fs/src/mem_fs/file.rs" ]
[]
true
wasmerio/wasmer
6,653
issue_to_patch
Symlinks webc
Added support for packages with symlinks (based on webc v12)
df7b69a3da75ced15ce2fee28cec72d59314a881
e3ec9729fba770c67c343973db2858edb0893f25
diff --git a/Cargo.lock b/Cargo.lock index d70fac0debb0..95385dfc919d 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -7876,9 +7876,9 @@ dependencies = [ [[package]] name = "webc" -version = "11.0.0" +version = "12.0.0" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "20b8b523112384e8c1caf...
[ "Cargo.lock", "Cargo.toml", "lib/cli/src/commands/create_exe.rs", "lib/package/src/package/package.rs", "lib/package/src/package/volume/fs.rs", "lib/package/src/package/volume/in_memory.rs", "lib/virtual-fs/src/webc_volume_fs.rs", "lib/wasix/src/fs/mod.rs" ]
[ { "comment": "`FsVolume::metadata()` currently returns `None` if `Timestamps::from_metadata(&meta)` fails. That causes entries (including symlinks) to effectively disappear from the packaged volume on platforms/filesystems where timestamp extraction isn’t supported or errors. Prefer returning metadata with `tim...
true
wasmerio/wasmer
6,653
comment_to_fix
Symlinks webc
`FsVolume::metadata()` currently returns `None` if `Timestamps::from_metadata(&meta)` fails. That causes entries (including symlinks) to effectively disappear from the packaged volume on platforms/filesystems where timestamp extraction isn’t supported or errors. Prefer returning metadata with `timestamps: None` instead...
df7b69a3da75ced15ce2fee28cec72d59314a881
e3ec9729fba770c67c343973db2858edb0893f25
diff --git a/lib/package/src/package/volume/fs.rs b/lib/package/src/package/volume/fs.rs index 6077a191a40b..3cb349148481 100644 --- a/lib/package/src/package/volume/fs.rs +++ b/lib/package/src/package/volume/fs.rs @@ -1,7 +1,7 @@ use std::{ collections::{BTreeMap, BTreeSet}, fmt::Debug, - fs::File, + ...
[ "lib/package/src/package/volume/fs.rs" ]
[ { "comment": "`FsVolume::metadata()` currently returns `None` if `Timestamps::from_metadata(&meta)` fails. That causes entries (including symlinks) to effectively disappear from the packaged volume on platforms/filesystems where timestamp extraction isn’t supported or errors. Prefer returning metadata with `tim...
true
wasmerio/wasmer
6,653
comment_to_fix
Symlinks webc
When serializing a volume to WebC v1, symlink entries are silently dropped. This can produce executables with missing files/dirs and no indication of why. If v1 can’t represent symlinks, emit at least a warning so users can diagnose the issue (or consider failing fast).
df7b69a3da75ced15ce2fee28cec72d59314a881
e3ec9729fba770c67c343973db2858edb0893f25
diff --git a/lib/cli/src/commands/create_exe.rs b/lib/cli/src/commands/create_exe.rs index 3001a216005e..743ddc0a0dee 100644 --- a/lib/cli/src/commands/create_exe.rs +++ b/lib/cli/src/commands/create_exe.rs @@ -541,6 +541,7 @@ fn serialize_volume_to_webc_v1(volume: &WebcVolume) -> Vec<u8> { ); ...
[ "lib/cli/src/commands/create_exe.rs" ]
[ { "comment": "When serializing a volume to WebC v1, symlink entries are silently dropped. This can produce executables with missing files/dirs and no indication of why. If v1 can’t represent symlinks, emit at least a warning so users can diagnose the issue (or consider failing fast).", "path": "lib/cli/src/...
true
wasmerio/wasmer
6,653
comment_to_fix
Symlinks webc
`FsVolume::metadata()` currently returns `None` if `Timestamps::from_metadata(&meta)` fails. That causes entries (including symlinks) to effectively disappear from the packaged volume on platforms/filesystems where timestamp extraction isn’t supported or errors. Prefer returning metadata with `timestamps: None` instead...
df7b69a3da75ced15ce2fee28cec72d59314a881
e3ec9729fba770c67c343973db2858edb0893f25
diff --git a/lib/package/src/package/volume/fs.rs b/lib/package/src/package/volume/fs.rs index 6077a191a40b..3cb349148481 100644 --- a/lib/package/src/package/volume/fs.rs +++ b/lib/package/src/package/volume/fs.rs @@ -1,7 +1,7 @@ use std::{ collections::{BTreeMap, BTreeSet}, fmt::Debug, - fs::File, + ...
[ "lib/package/src/package/volume/fs.rs" ]
[ { "comment": "`FsVolume::metadata()` currently returns `None` if `Timestamps::from_metadata(&meta)` fails. That causes entries (including symlinks) to effectively disappear from the packaged volume on platforms/filesystems where timestamp extraction isn’t supported or errors. Prefer returning metadata with `tim...
true
wasmerio/wasmer
6,653
comment_to_fix
Symlinks webc
When serializing a volume to WebC v1, symlink entries are silently dropped. This can produce executables with missing files/dirs and no indication of why. If v1 can’t represent symlinks, emit at least a warning so users can diagnose the issue (or consider failing fast).
df7b69a3da75ced15ce2fee28cec72d59314a881
e3ec9729fba770c67c343973db2858edb0893f25
diff --git a/lib/cli/src/commands/create_exe.rs b/lib/cli/src/commands/create_exe.rs index 3001a216005e..743ddc0a0dee 100644 --- a/lib/cli/src/commands/create_exe.rs +++ b/lib/cli/src/commands/create_exe.rs @@ -541,6 +541,7 @@ fn serialize_volume_to_webc_v1(volume: &WebcVolume) -> Vec<u8> { ); ...
[ "lib/cli/src/commands/create_exe.rs" ]
[ { "comment": "When serializing a volume to WebC v1, symlink entries are silently dropped. This can produce executables with missing files/dirs and no indication of why. If v1 can’t represent symlinks, emit at least a warning so users can diagnose the issue (or consider failing fast).", "path": "lib/cli/src/...
true
wasmerio/wasmer
6,647
issue_to_patch
Refactor duplicated logic to `function_extents()` helper
A small follow-up to https://github.com/wasmerio/wasmer/pull/6603: moving duplicated logic between `internal_register_frame_info` and `finished_function_extents` into a local helper on `AllocatedArtifact`. This also ensures that `internal_register_frame_info` has the same length matching assertion.
7ea276ea9f3c886b5c759eeccdb3f667db5bf078
65cc3f04a2c1ade600a9a3d247adbb5df3d72f07
diff --git a/lib/compiler/src/engine/artifact.rs b/lib/compiler/src/engine/artifact.rs index f2be6afe9d49..a19c9a3ff049 100644 --- a/lib/compiler/src/engine/artifact.rs +++ b/lib/compiler/src/engine/artifact.rs @@ -67,6 +67,23 @@ pub struct AllocatedArtifact { finished_function_lengths: BoxedSlice<LocalFunctionInd...
[ "lib/compiler/src/engine/artifact.rs" ]
[]
true
wasmerio/wasmer
6,626
issue_to_patch
Fix WASIX symlink resolution for path components
This fixes WASIX path resolution so symlinks are handled consistently during component traversal. Changes: - Resolve intermediate symlink components even when the final component should not be followed. - Follow the final symlink only when `follow_symlinks` is enabled. - Cache backing filesystem symlinks as `Ki...
d529bf7f901f50f8858c3808d8fe385bd6093acf
1031ba8ce884d5f2d1a5d38d7cd4bf262cfba2f1
diff --git a/lib/virtual-fs/src/arc_fs.rs b/lib/virtual-fs/src/arc_fs.rs index 472985a0ed10..d3ca55c82b09 100644 --- a/lib/virtual-fs/src/arc_fs.rs +++ b/lib/virtual-fs/src/arc_fs.rs @@ -38,6 +38,10 @@ impl FileSystem for ArcFileSystem { self.fs.create_symlink(source, target) } + fn hard_link(&self, ...
[ "lib/virtual-fs/src/arc_fs.rs", "lib/virtual-fs/src/host_fs.rs", "lib/virtual-fs/src/lib.rs", "lib/virtual-fs/src/mount_fs.rs", "lib/virtual-fs/src/overlay_fs.rs", "lib/virtual-fs/src/passthru_fs.rs", "lib/virtual-fs/src/tmp_fs.rs", "lib/wasix/src/fs/fd.rs", "lib/wasix/src/fs/mod.rs", "lib/wasix/s...
[ { "comment": "For readability, let's declare that outside the function.", "path": "lib/wasix/src/fs/mod.rs", "hunk": "@@ -1135,101 +1135,147 @@ impl WasiFs {\n // assume '/a/b' to be the same as `a/b` relative to the inode, so\n // we skip over the RootDir component.\n ...
true
wasmerio/wasmer
6,626
comment_to_fix
Fix WASIX symlink resolution for path components
For readability, let's declare that outside the function.
d529bf7f901f50f8858c3808d8fe385bd6093acf
1031ba8ce884d5f2d1a5d38d7cd4bf262cfba2f1
diff --git a/lib/wasix/src/fs/mod.rs b/lib/wasix/src/fs/mod.rs index 8e83a0f2b2f9..c64cbb6b1b75 100644 --- a/lib/wasix/src/fs/mod.rs +++ b/lib/wasix/src/fs/mod.rs @@ -18,12 +18,13 @@ mod fd; mod fd_list; mod inode_guard; mod notification; +mod path_posix; use std::{ - borrow::{Borrow, Cow}, + borrow::Cow, ...
[ "lib/wasix/src/fs/mod.rs" ]
[ { "comment": "For readability, let's declare that outside the function.", "path": "lib/wasix/src/fs/mod.rs", "hunk": "@@ -1135,101 +1135,147 @@ impl WasiFs {\n // assume '/a/b' to be the same as `a/b` relative to the inode, so\n // we skip over the RootDir component.\n ...
true
wasmerio/wasmer
6,626
comment_to_fix
Fix WASIX symlink resolution for path components
This should be changed to: `follow_symlinks || !last_component`, because for /A/B/C where B->Z and C->Y: - follow=true => /Z/Y - follow=false => /Z/C So for B we pass `!last_component` (always true), and for C `follow_symlinks`.
d529bf7f901f50f8858c3808d8fe385bd6093acf
1031ba8ce884d5f2d1a5d38d7cd4bf262cfba2f1
diff --git a/lib/wasix/src/fs/mod.rs b/lib/wasix/src/fs/mod.rs index 8e83a0f2b2f9..c64cbb6b1b75 100644 --- a/lib/wasix/src/fs/mod.rs +++ b/lib/wasix/src/fs/mod.rs @@ -18,12 +18,13 @@ mod fd; mod fd_list; mod inode_guard; mod notification; +mod path_posix; use std::{ - borrow::{Borrow, Cow}, + borrow::Cow, ...
[ "lib/wasix/src/fs/mod.rs" ]
[ { "comment": "This should be changed to: `follow_symlinks || !last_component`, because for /A/B/C where B->Z and C->Y:\n\n- follow=true => /Z/Y\n- follow=false => /Z/C\n\nSo for B we pass `!last_component` (always true), and for C `follow_symlinks`.", "path": "lib/wasix/src/fs/mod.rs", "hunk": "@@ -1272...
true
wasmerio/wasmer
6,626
comment_to_fix
Fix WASIX symlink resolution for path components
`entry_names.contains(*name)` attempts to move out of `&String` (and won’t compile). Use `contains(name)` instead to compare by reference.
d529bf7f901f50f8858c3808d8fe385bd6093acf
1031ba8ce884d5f2d1a5d38d7cd4bf262cfba2f1
diff --git a/lib/wasix/src/syscalls/wasi/fd_readdir.rs b/lib/wasix/src/syscalls/wasi/fd_readdir.rs index 5672ab5f6ff6..7a351a1cb844 100644 --- a/lib/wasix/src/syscalls/wasi/fd_readdir.rs +++ b/lib/wasix/src/syscalls/wasi/fd_readdir.rs @@ -66,16 +66,17 @@ pub fn fd_readdir<M: MemorySize>( }) ...
[ "lib/wasix/src/syscalls/wasi/fd_readdir.rs" ]
[ { "comment": "`entry_names.contains(*name)` attempts to move out of `&String` (and won’t compile). Use `contains(name)` instead to compare by reference.", "path": "lib/wasix/src/syscalls/wasi/fd_readdir.rs", "hunk": "@@ -66,16 +66,17 @@ pub fn fd_readdir<M: MemorySize>(\n })\n ...
true
wasmerio/wasmer
6,626
comment_to_fix
Fix WASIX symlink resolution for path components
`relative_path_to_absolute` turns an empty pathname into a non-empty `cwd/` path, which can make `chdir("")` succeed instead of returning `Errno::Noent`. Reject the empty pathname before absolutizing so `chdir` matches the resolver’s empty-path behavior.
d529bf7f901f50f8858c3808d8fe385bd6093acf
1031ba8ce884d5f2d1a5d38d7cd4bf262cfba2f1
diff --git a/lib/wasix/src/syscalls/wasix/chdir.rs b/lib/wasix/src/syscalls/wasix/chdir.rs index e7e5018d4d6c..6f120a381d39 100644 --- a/lib/wasix/src/syscalls/wasix/chdir.rs +++ b/lib/wasix/src/syscalls/wasix/chdir.rs @@ -30,14 +30,59 @@ pub fn chdir<M: MemorySize>( Ok(Errno::Success) } +/// Change the WASIX c...
[ "lib/wasix/src/syscalls/wasix/chdir.rs" ]
[ { "comment": "`relative_path_to_absolute` turns an empty pathname into a non-empty `cwd/` path, which can make `chdir(\"\")` succeed instead of returning `Errno::Noent`. Reject the empty pathname before absolutizing so `chdir` matches the resolver’s empty-path behavior.", "path": "lib/wasix/src/syscalls/was...
true
wasmerio/wasmer
6,626
comment_to_fix
Fix WASIX symlink resolution for path components
The host FS should not support creating links to avoid sandbox escape scenarios.
d529bf7f901f50f8858c3808d8fe385bd6093acf
1031ba8ce884d5f2d1a5d38d7cd4bf262cfba2f1
diff --git a/lib/virtual-fs/src/host_fs.rs b/lib/virtual-fs/src/host_fs.rs index d7fff96808bb..1c51ef5d9fc8 100644 --- a/lib/virtual-fs/src/host_fs.rs +++ b/lib/virtual-fs/src/host_fs.rs @@ -67,6 +67,40 @@ pub fn normalize_path(path: &Path) -> PathBuf { ret } +fn path_suffix_to_guest_absolute(stripped: &Path) -...
[ "lib/virtual-fs/src/host_fs.rs" ]
[ { "comment": "The host FS should not support creating links to avoid sandbox escape scenarios.", "path": "lib/virtual-fs/src/host_fs.rs", "hunk": "@@ -138,6 +138,17 @@ impl crate::FileSystem for FileSystem {\n fs::create_dir(path).map_err(Into::into)\n }\n \n+ fn hard_link(&self, source: ...
true
wasmerio/wasmer
6,626
comment_to_fix
Fix WASIX symlink resolution for path components
Why we need this function? Is not this one better? (it doesn't require string lossy conversion) ```suggestion path.is_absolute() ```
d529bf7f901f50f8858c3808d8fe385bd6093acf
1031ba8ce884d5f2d1a5d38d7cd4bf262cfba2f1
diff --git a/lib/wasix/src/fs/mod.rs b/lib/wasix/src/fs/mod.rs index 8e83a0f2b2f9..c64cbb6b1b75 100644 --- a/lib/wasix/src/fs/mod.rs +++ b/lib/wasix/src/fs/mod.rs @@ -18,12 +18,13 @@ mod fd; mod fd_list; mod inode_guard; mod notification; +mod path_posix; use std::{ - borrow::{Borrow, Cow}, + borrow::Cow, ...
[ "lib/wasix/src/fs/mod.rs" ]
[ { "comment": "Why we need this function? \n\nIs not this one better? (it doesn't require string lossy conversion)\n\n```suggestion\n path.is_absolute()\n```", "path": "lib/wasix/src/fs/mod.rs", "hunk": "@@ -302,26 +331,181 @@ pub struct WasiInodes {\n protected: Arc<RwLock<WasiInodesProtected>>,\...
true
wasmerio/wasmer
6,626
comment_to_fix
Fix WASIX symlink resolution for path components
Why not this? ```suggestion PathBuf::from(path.strip_prefix("/").unwrap_or(path)) ```
d529bf7f901f50f8858c3808d8fe385bd6093acf
1031ba8ce884d5f2d1a5d38d7cd4bf262cfba2f1
diff --git a/lib/wasix/src/fs/mod.rs b/lib/wasix/src/fs/mod.rs index 8e83a0f2b2f9..c64cbb6b1b75 100644 --- a/lib/wasix/src/fs/mod.rs +++ b/lib/wasix/src/fs/mod.rs @@ -18,12 +18,13 @@ mod fd; mod fd_list; mod inode_guard; mod notification; +mod path_posix; use std::{ - borrow::{Borrow, Cow}, + borrow::Cow, ...
[ "lib/wasix/src/fs/mod.rs" ]
[ { "comment": "Why not this?\n\n```suggestion\n PathBuf::from(path.strip_prefix(\"/\").unwrap_or(path))\n```", "path": "lib/wasix/src/fs/mod.rs", "hunk": "@@ -302,26 +331,181 @@ pub struct WasiInodes {\n protected: Arc<RwLock<WasiInodesProtected>>,\n }\n \n+// Normalize only already-resolved virtu...
true
wasmerio/wasmer
6,644
issue_to_patch
fix(V8): use proper new memory sharing mechanism (take 2)
This is a second integration attempt for the V8 back-end where we properly distinguish between a memory attached to a Store (`VMMemory`) and a shared one, `VMSharedMemory` that can be later used in a different `Store`. With that, we can properly shared and access memory in between different Isolates. From the implement...
caf909f12ce8124141fb1d32c12c51ad8b335e20
d4d79a2040e9ab831e129432e49478f440427f4a
diff --git a/Cargo.toml b/Cargo.toml index 63e5fffa5119..585eb59aa176 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -106,7 +106,7 @@ axum = { version = "=0.6.9", default-features = false } backtrace = "0.3" base64 = "0.22.1" bincode = "=2.0.1" -bindgen = "0.72.1" +bindgen = { version = "0.72.1", default-features = fal...
[ "Cargo.toml", "Makefile", "lib/api/Cargo.toml", "lib/api/build.rs", "lib/api/src/backend/js/entities/memory/mod.rs", "lib/api/src/backend/sys/entities/memory/mod.rs", "lib/api/src/backend/v8/entities/memory/mod.rs", "lib/api/src/backend/v8/vm/mod.rs", "lib/api/src/entities/memory/inner.rs", "lib/a...
[ { "comment": "This V8 method invokes `wasm_memory_type`/`wasm_memory_share` through `self.handle.as_shared()` without first checking that the caller is on the store's isolate thread. Other V8 memory methods call `check_isolate(store)` before touching the V8 C API; skipping it here can call into V8 from the wron...
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index d64104af823c..effddb68b19e 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -931,9 +931,8 @@ fn collect_tests(tests: &mut Vec<Trial>) -> Result<()> { let mut supported_engines = v...
true
wasmerio/wasmer
6,644
comment_to_fix
fix(V8): use proper new memory sharing mechanism (take 2)
This V8 method invokes `wasm_memory_type`/`wasm_memory_share` through `self.handle.as_shared()` without first checking that the caller is on the store's isolate thread. Other V8 memory methods call `check_isolate(store)` before touching the V8 C API; skipping it here can call into V8 from the wrong thread when `Memory:...
caf909f12ce8124141fb1d32c12c51ad8b335e20
d4d79a2040e9ab831e129432e49478f440427f4a
diff --git a/lib/api/src/backend/v8/entities/memory/mod.rs b/lib/api/src/backend/v8/entities/memory/mod.rs index 8d876298a120..ebc79a6b5c79 100644 --- a/lib/api/src/backend/v8/entities/memory/mod.rs +++ b/lib/api/src/backend/v8/entities/memory/mod.rs @@ -162,10 +162,62 @@ impl Memory { pub fn copy(&self, store: &i...
[ "lib/api/src/backend/v8/entities/memory/mod.rs" ]
[ { "comment": "This V8 method invokes `wasm_memory_type`/`wasm_memory_share` through `self.handle.as_shared()` without first checking that the caller is on the store's isolate thread. Other V8 memory methods call `check_isolate(store)` before touching the V8 C API; skipping it here can call into V8 from the wron...
true
wasmerio/wasmer
6,644
comment_to_fix
fix(V8): use proper new memory sharing mechanism (take 2)
This uses the build script's host `cfg!(target_os)` instead of the Cargo target OS. When cross-compiling V8 for Windows from a non-Windows host, the script will look for `libv8.a` even though the selected Windows archive contains `v8.lib`. Use `CARGO_CFG_TARGET_OS` consistently here, as the asset selection above does.
caf909f12ce8124141fb1d32c12c51ad8b335e20
d4d79a2040e9ab831e129432e49478f440427f4a
diff --git a/lib/api/build.rs b/lib/api/build.rs index 4dc1ef9963fa..354c05fb646f 100644 --- a/lib/api/build.rs +++ b/lib/api/build.rs @@ -18,10 +18,9 @@ fn build_v8() { ) { ("macos", "aarch64", _) => ("v8-darwin-aarch64.tar.xz", "darwin-aarch64"), ("linux", "x86_64", "gnu") => ("v8-linux-amd64.t...
[ "lib/api/build.rs" ]
[ { "comment": "This uses the build script's host `cfg!(target_os)` instead of the Cargo target OS. When cross-compiling V8 for Windows from a non-Windows host, the script will look for `libv8.a` even though the selected Windows archive contains `v8.lib`. Use `CARGO_CFG_TARGET_OS` consistently here, as the asset ...
true
wasmerio/wasmer
6,644
comment_to_fix
fix(V8): use proper new memory sharing mechanism (take 2)
Typo in the comment: "workardound" should be "workaround".
caf909f12ce8124141fb1d32c12c51ad8b335e20
d4d79a2040e9ab831e129432e49478f440427f4a
diff --git a/scripts/alpine-linux-install-deps.sh b/scripts/alpine-linux-install-deps.sh index 5a88cb018a83..fcdd95a0172c 100755 --- a/scripts/alpine-linux-install-deps.sh +++ b/scripts/alpine-linux-install-deps.sh @@ -9,5 +9,17 @@ apk update # build scripts (proc-macro2, quote, libc, serde_core, ...). Without it, # ...
[ "scripts/alpine-linux-install-deps.sh" ]
[ { "comment": "Typo in the comment: \"workardound\" should be \"workaround\".", "path": "scripts/alpine-linux-install-deps.sh", "hunk": "@@ -10,4 +10,16 @@ apk update\n # `cargo build` fails with `error: linker 'cc' not found` on alpine:edge\n # images that no longer ship a host C toolchain by default.\n...
true
wasmerio/wasmer
6,644
comment_to_fix
fix(V8): use proper new memory sharing mechanism (take 2)
Typo in the comment: "must" should be "musl".
caf909f12ce8124141fb1d32c12c51ad8b335e20
d4d79a2040e9ab831e129432e49478f440427f4a
diff --git a/lib/api/Cargo.toml b/lib/api/Cargo.toml index 77b9f76d3f90..dd3311075280 100644 --- a/lib/api/Cargo.toml +++ b/lib/api/Cargo.toml @@ -185,7 +185,17 @@ ureq = { workspace = true, optional = true } which = { workspace = true, optional = true } xz = { workspace = true, optional = true } zip = { workspace =...
[ "lib/api/Cargo.toml" ]
[ { "comment": "Typo in the comment: \"must\" should be \"musl\".", "path": "lib/api/Cargo.toml", "hunk": "@@ -185,7 +185,17 @@ ureq = { workspace = true, optional = true }\n which = { workspace = true, optional = true }\n xz = { workspace = true, optional = true }\n zip = { workspace = true, optional = t...
true
wasmerio/wasmer
6,644
comment_to_fix
fix(V8): use proper new memory sharing mechanism (take 2)
Typo in the comment: "workardound" should be "workaround".
caf909f12ce8124141fb1d32c12c51ad8b335e20
d4d79a2040e9ab831e129432e49478f440427f4a
diff --git a/scripts/alpine-linux-install-deps.sh b/scripts/alpine-linux-install-deps.sh index 5a88cb018a83..fcdd95a0172c 100755 --- a/scripts/alpine-linux-install-deps.sh +++ b/scripts/alpine-linux-install-deps.sh @@ -9,5 +9,17 @@ apk update # build scripts (proc-macro2, quote, libc, serde_core, ...). Without it, # ...
[ "scripts/alpine-linux-install-deps.sh" ]
[ { "comment": "Typo in the comment: \"workardound\" should be \"workaround\".", "path": "scripts/alpine-linux-install-deps.sh", "hunk": "@@ -10,4 +10,16 @@ apk update\n # `cargo build` fails with `error: linker 'cc' not found` on alpine:edge\n # images that no longer ship a host C toolchain by default.\n...
true
wasmerio/wasmer
6,644
comment_to_fix
fix(V8): use proper new memory sharing mechanism (take 2)
Typo in the comment: "must" should be "musl".
caf909f12ce8124141fb1d32c12c51ad8b335e20
d4d79a2040e9ab831e129432e49478f440427f4a
diff --git a/lib/api/Cargo.toml b/lib/api/Cargo.toml index 77b9f76d3f90..dd3311075280 100644 --- a/lib/api/Cargo.toml +++ b/lib/api/Cargo.toml @@ -185,7 +185,17 @@ ureq = { workspace = true, optional = true } which = { workspace = true, optional = true } xz = { workspace = true, optional = true } zip = { workspace =...
[ "lib/api/Cargo.toml" ]
[ { "comment": "Typo in the comment: \"must\" should be \"musl\".", "path": "lib/api/Cargo.toml", "hunk": "@@ -185,7 +185,17 @@ ureq = { workspace = true, optional = true }\n which = { workspace = true, optional = true }\n xz = { workspace = true, optional = true }\n zip = { workspace = true, optional = t...
true
wasmerio/wasmer
6,627
issue_to_patch
fix: advance static WebC file cursors on read
## What Static WebC reads in `StaticFileSystem` were not moving the file cursor. If a caller reads the file in chunks, it gets the same bytes again, which is pretty annoying. This path is live in the C API WASI `prepare_webc_env` flow too, so this is not just test-only stuff. ## Repro Use any static WebC file that is...
e6cf863f4198a4d40254ebe754fc2635a552f1ce
a2d2c3223e009c18b4b86b48420003b26404b129
diff --git a/lib/virtual-fs/src/static_fs.rs b/lib/virtual-fs/src/static_fs.rs index 365829431c43..67ab796ae6f9 100644 --- a/lib/virtual-fs/src/static_fs.rs +++ b/lib/virtual-fs/src/static_fs.rs @@ -135,27 +135,27 @@ impl AsyncRead for WebCFile { _cx: &mut Context<'_>, buf: &mut tokio::io::ReadBuf<'_>...
[ "lib/virtual-fs/src/static_fs.rs" ]
[ { "comment": "Let's stick to the usage of `self`, introduction of `this` seem not idiomatic to Rust.", "path": "lib/virtual-fs/src/static_fs.rs", "hunk": "@@ -135,27 +135,27 @@ impl AsyncRead for WebCFile {\n _cx: &mut Context<'_>,\n buf: &mut tokio::io::ReadBuf<'_>,\n ) -> Poll<io::...
true
wasmerio/wasmer
6,593
issue_to_patch
chore: fix duplicated words in wasix and vm doc comments
Three small comment-only typos: doubled `a a` in `lib/wasix/src/os/task/process.rs` (two lines) and `lib/wasix/src/runtime/mod.rs`. No functional change.
4a7d25af9d8cd1ea8128424ed2042ebcfaf66d3e
abb23ed53509f721a24274249da762cf6ee559b6
diff --git a/lib/wasix/src/os/task/process.rs b/lib/wasix/src/os/task/process.rs index 25528b1d75f9..2baf7cbfebe5 100644 --- a/lib/wasix/src/os/task/process.rs +++ b/lib/wasix/src/os/task/process.rs @@ -497,7 +497,7 @@ impl WasiProcess { self.inner.0.lock().unwrap() } - /// Creates a a thread and ret...
[ "lib/wasix/src/os/task/process.rs", "lib/wasix/src/runtime/mod.rs" ]
[]
true
wasmerio/wasmer
6,642
issue_to_patch
Fix FS race resulting in None handle, take 2
Supersedes #6636.
caf909f12ce8124141fb1d32c12c51ad8b335e20
a14df89a8c83f760c76ed41930cd668daaa412bd
diff --git a/lib/wasix/src/fs/fd.rs b/lib/wasix/src/fs/fd.rs index 5fe42b2f1f5c..0c018aaee852 100644 --- a/lib/wasix/src/fs/fd.rs +++ b/lib/wasix/src/fs/fd.rs @@ -15,6 +15,9 @@ use crate::os::epoll::EpollState; use super::{InodeGuard, InodeWeakGuard, NotificationInner}; +/// Shared handle to an open [`VirtualFile`...
[ "lib/wasix/src/fs/fd.rs", "lib/wasix/src/fs/fd_list.rs", "lib/wasix/src/fs/inode_guard.rs", "lib/wasix/src/fs/mod.rs", "lib/wasix/src/journal/effector/syscalls/fd_close.rs", "lib/wasix/src/syscalls/wasi/fd_close.rs", "lib/wasix/src/syscalls/wasi/fd_datasync.rs", "lib/wasix/src/syscalls/wasi/fd_fdstat_...
[ { "comment": "`dup2_at`’s doc comment says the flush target must be captured while holding `fd_map.write()` and before `drop_one_handle` can clear the inode’s `Kind::File { handle: None }`. However, the current implementation removes `dst` first (which can drop the last handle and clear `handle`), and only then...
diff --git a/lib/wasix/tests/wasm_tests/fd_tests/fd-dup2-same-inode/main.c b/lib/wasix/tests/wasm_tests/fd_tests/fd-dup2-same-inode/main.c new file mode 100644 index 000000000000..6e9266743237 --- /dev/null +++ b/lib/wasix/tests/wasm_tests/fd_tests/fd-dup2-same-inode/main.c @@ -0,0 +1,93 @@ +//#ExpectedStdout: dup2 sam...
true
wasmerio/wasmer
6,642
comment_to_fix
Fix FS race resulting in None handle, take 2
`dup2_at`’s doc comment says the flush target must be captured while holding `fd_map.write()` and before `drop_one_handle` can clear the inode’s `Kind::File { handle: None }`. However, the current implementation removes `dst` first (which can drop the last handle and clear `handle`), and only then derives the flush tar...
caf909f12ce8124141fb1d32c12c51ad8b335e20
a14df89a8c83f760c76ed41930cd668daaa412bd
diff --git a/lib/wasix/src/fs/mod.rs b/lib/wasix/src/fs/mod.rs index be028f20929d..8e83a0f2b2f9 100644 --- a/lib/wasix/src/fs/mod.rs +++ b/lib/wasix/src/fs/mod.rs @@ -4,6 +4,15 @@ // only way we can get to a file on a FileSystem instance is by going // through its respective FileOpener and giving it a path as input. ...
[ "lib/wasix/src/fs/mod.rs" ]
[ { "comment": "`dup2_at`’s doc comment says the flush target must be captured while holding `fd_map.write()` and before `drop_one_handle` can clear the inode’s `Kind::File { handle: None }`. However, the current implementation removes `dst` first (which can drop the last handle and clear `handle`), and only then...
true
wasmerio/wasmer
6,642
comment_to_fix
Fix FS race resulting in None handle, take 2
`TARGET_FD` is intentionally kept open in the parent to ensure the spawned child’s dup2 replaces an occupied target, but it is never closed in the parent afterward. Since `posix_spawn_file_actions_adddup2` only affects the child, this leaks the parent’s FD 10 and can make the later `unlink(decoy_path)` cleanup dependen...
caf909f12ce8124141fb1d32c12c51ad8b335e20
a14df89a8c83f760c76ed41930cd668daaa412bd
diff --git a/lib/wasix/tests/wasm_tests/fd_tests/proc-spawn2-dup2/main.c b/lib/wasix/tests/wasm_tests/fd_tests/proc-spawn2-dup2/main.c new file mode 100644 index 000000000000..8e7db49b72b7 --- /dev/null +++ b/lib/wasix/tests/wasm_tests/fd_tests/proc-spawn2-dup2/main.c @@ -0,0 +1,105 @@ +//#ExpectedStdout: proc_spawn2 d...
[ "lib/wasix/tests/wasm_tests/fd_tests/proc-spawn2-dup2/main.c" ]
[ { "comment": "`TARGET_FD` is intentionally kept open in the parent to ensure the spawned child’s dup2 replaces an occupied target, but it is never closed in the parent afterward. Since `posix_spawn_file_actions_adddup2` only affects the child, this leaks the parent’s FD 10 and can make the later `unlink(decoy_p...
true
wasmerio/wasmer
6,642
comment_to_fix
Fix FS race resulting in None handle, take 2
`pthread_*` APIs return an error code and generally do not set `errno`, so using `perror()` here can print a misleading message (sometimes even "Success") and loses the actual failure reason. Capture the return code and print `strerror(rc)` instead for barrier init and thread creation.
caf909f12ce8124141fb1d32c12c51ad8b335e20
a14df89a8c83f760c76ed41930cd668daaa412bd
diff --git a/lib/wasix/tests/wasm_tests/fd_tests/fd-read-none-handle-race/main.c b/lib/wasix/tests/wasm_tests/fd_tests/fd-read-none-handle-race/main.c new file mode 100644 index 000000000000..d9a215f72e9b --- /dev/null +++ b/lib/wasix/tests/wasm_tests/fd_tests/fd-read-none-handle-race/main.c @@ -0,0 +1,175 @@ +//#Expec...
[ "lib/wasix/tests/wasm_tests/fd_tests/fd-read-none-handle-race/main.c" ]
[ { "comment": "`pthread_*` APIs return an error code and generally do not set `errno`, so using `perror()` here can print a misleading message (sometimes even \"Success\") and loses the actual failure reason. Capture the return code and print `strerror(rc)` instead for barrier init and thread creation.", "pa...
true
wasmerio/wasmer
6,642
comment_to_fix
Fix FS race resulting in None handle, take 2
Same issue as earlier: `pthread_join` / `pthread_barrier_destroy` return an error code rather than setting `errno`, so `perror()` can’t reliably report failures. Use the return code with `strerror(rc)` so failures are actionable.
caf909f12ce8124141fb1d32c12c51ad8b335e20
a14df89a8c83f760c76ed41930cd668daaa412bd
diff --git a/lib/wasix/tests/wasm_tests/fd_tests/fd-read-none-handle-race/main.c b/lib/wasix/tests/wasm_tests/fd_tests/fd-read-none-handle-race/main.c new file mode 100644 index 000000000000..d9a215f72e9b --- /dev/null +++ b/lib/wasix/tests/wasm_tests/fd_tests/fd-read-none-handle-race/main.c @@ -0,0 +1,175 @@ +//#Expec...
[ "lib/wasix/tests/wasm_tests/fd_tests/fd-read-none-handle-race/main.c" ]
[ { "comment": "Same issue as earlier: `pthread_join` / `pthread_barrier_destroy` return an error code rather than setting `errno`, so `perror()` can’t reliably report failures. Use the return code with `strerror(rc)` so failures are actionable.", "path": "lib/wasix/tests/wasm_tests/fd_tests/fd-read-none-hand...
true
wasmerio/wasmer
6,642
comment_to_fix
Fix FS race resulting in None handle, take 2
Java vibes, double boxing , not so great for perf. *if* we never need to move out or swap the file, we can use: `Arc<RwLock<dyn VirtualFile>>` without the extra Box, I believe.
caf909f12ce8124141fb1d32c12c51ad8b335e20
a14df89a8c83f760c76ed41930cd668daaa412bd
diff --git a/lib/wasix/src/fs/fd.rs b/lib/wasix/src/fs/fd.rs index 5fe42b2f1f5c..0c018aaee852 100644 --- a/lib/wasix/src/fs/fd.rs +++ b/lib/wasix/src/fs/fd.rs @@ -15,6 +15,9 @@ use crate::os::epoll::EpollState; use super::{InodeGuard, InodeWeakGuard, NotificationInner}; +/// Shared handle to an open [`VirtualFile`...
[ "lib/wasix/src/fs/fd.rs" ]
[ { "comment": "Java vibes, double boxing , not so great for perf.\n\n*if* we never need to move out or swap the file, we can use: `Arc<RwLock<dyn VirtualFile>>` without the extra Box, I believe.", "path": "lib/wasix/src/fs/fd.rs", "hunk": "@@ -15,6 +15,9 @@ use crate::os::epoll::EpollState;\n \n use supe...
true
wasmerio/wasmer
6,642
comment_to_fix
Fix FS race resulting in None handle, take 2
Side question, but why Option<>. So we can retain info for a closed fd?
caf909f12ce8124141fb1d32c12c51ad8b335e20
a14df89a8c83f760c76ed41930cd668daaa412bd
diff --git a/lib/wasix/src/fs/fd.rs b/lib/wasix/src/fs/fd.rs index 5fe42b2f1f5c..0c018aaee852 100644 --- a/lib/wasix/src/fs/fd.rs +++ b/lib/wasix/src/fs/fd.rs @@ -15,6 +15,9 @@ use crate::os::epoll::EpollState; use super::{InodeGuard, InodeWeakGuard, NotificationInner}; +/// Shared handle to an open [`VirtualFile`...
[ "lib/wasix/src/fs/fd.rs" ]
[ { "comment": "Side question, but why Option<>.\r\nSo we can retain info for a closed fd?", "path": "lib/wasix/src/fs/fd.rs", "hunk": "@@ -89,7 +92,7 @@ pub enum Kind {\n File {\n /// The open file, if it's open\n #[cfg_attr(feature = \"enable-serde\", serde(skip))]\n- handle: ...
true
wasmerio/wasmer
6,642
comment_to_fix
Fix FS race resulting in None handle, take 2
I don't think this is new, it was already broken before, but fd_renumber is supposed to just RENUMBER, not to keep the old fd open! But since this is not new behaviour, can come as a followup.
caf909f12ce8124141fb1d32c12c51ad8b335e20
a14df89a8c83f760c76ed41930cd668daaa412bd
diff --git a/lib/wasix/src/syscalls/wasi/fd_renumber.rs b/lib/wasix/src/syscalls/wasi/fd_renumber.rs index ad7ae56858e1..6c42a18e37a4 100644 --- a/lib/wasix/src/syscalls/wasi/fd_renumber.rs +++ b/lib/wasix/src/syscalls/wasi/fd_renumber.rs @@ -1,5 +1,5 @@ use super::*; -use crate::fs::{FlushPoller, MAX_FD}; +use crate:...
[ "lib/wasix/src/syscalls/wasi/fd_renumber.rs" ]
[ { "comment": "I don't think this is new, it was already broken before, but fd_renumber is supposed to just RENUMBER, not to keep the old fd open!\n\nBut since this is not new behaviour, can come as a followup.", "path": "lib/wasix/src/syscalls/wasi/fd_renumber.rs", "hunk": "@@ -38,71 +38,15 @@ pub(crate...
true
wasmerio/wasmer
6,643
issue_to_patch
ci(alpine): install build-base so rustc can link host build scripts
The C-API (linux-musl) and test-wasmer-cli (linux-musl) jobs both fail on `alpine:edge` with: error: linker `cc` not found could not compile `quote` (build script) could not compile `proc-macro2` (build script) could not compile `libc` (build script) could not compile `serde_core` (build scri...
176021becfb1aef83a617f85ac549c116e16950d
b439cc33d1c88f914ff40f3a76113bde13c76138
diff --git a/scripts/alpine-linux-install-deps.sh b/scripts/alpine-linux-install-deps.sh index 841881eb0298..5a88cb018a83 100755 --- a/scripts/alpine-linux-install-deps.sh +++ b/scripts/alpine-linux-install-deps.sh @@ -4,5 +4,10 @@ # This script is used by the CI! apk update -apk add bash make curl cmake ninja clan...
[ "scripts/alpine-linux-install-deps.sh" ]
[]
true
wasmerio/wasmer
6,625
issue_to_patch
port wasi-fyi to wasmer-wasix tests I think the tests are pretty similar to the WASIX tests: - an input file (`rustc +nightly --target=wasm32-wasip1 -o "$output" "$input"`) is compiled to `.wasm` file - a test is run with a provided `stdin` - and expected `stdout`/`stderr` are sometimes expected @Arshia001 is it some...
test(wasix): port wasi-fyi suite to wasm harness
This ports the complete wasi-fyi fixture suite into the shared wasmer-wasix wasm_tests harness, so the fixtures are discovered as normal directive-driven wasm tests and run per engine through libtest-mimic. The shared harness now supports Rust primary sources, runtime env/stdin directives, stderr and file-backed outpu...
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/.github/workflows/test.yaml b/.github/workflows/test.yaml index 6b19276e787b..226eb23c1c13 100644 --- a/.github/workflows/test.yaml +++ b/.github/workflows/test.yaml @@ -149,51 +149,6 @@ jobs: - name: make test-js run: | make test-js - test_wasi_fyi: - name: Wasi-fyi - run...
[ ".github/workflows/test.yaml", "Makefile", "lib/wasix/tests/wasm_tests/README.md", "lib/wasix/tests/wasm_tests/mod.rs", "lib/wasix/tests/wasm_tests/runner.rs", "lib/wasix/tests/wasm_tests/wasi_fyi/LICENSE", "lib/wasix/tests/wasm_tests/wasi_fyi/README.md", "lib/wasix/tests/wasm_tests/wasi_fyi/env_args-...
[ { "comment": "`unwrap_or_default()` treats an unknown exit code as `0`, which can incorrectly pass fixtures that expect success (`*.status` contains `0`) even if the run actually failed but didn’t produce a parsable exit code. Prefer requiring `result.exit_code` to be `Some` when a `.status` sidecar exists (e.g...
diff --git a/lib/wasix/tests/wasm_tests/README.md b/lib/wasix/tests/wasm_tests/README.md index d50561813185..0a0d19bfd262 100644 --- a/lib/wasix/tests/wasm_tests/README.md +++ b/lib/wasix/tests/wasm_tests/README.md @@ -18,11 +18,14 @@ The harness builds each discovered fixture, runs the resulting `main` module thr and...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
Do we know why is the previous variant not sufficient?
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index a4d5abd2fec4..d64104af823c 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -4,8 +4,8 @@ //! for a WASM test. Each configuration represents a distinct test run, with its //! own arguments,...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "Do we know why is the previous variant not sufficient?", "path": "lib/wasix/tests/wasm_tests/mod.rs", "hunk": "@@ -384,13 +461,11 @@ fn run_build_script(config: &Config) -> anyhow::Result<PathBuf> {\n }\n create_dir_all(&build_test_path)?;\n \n- let mut options = fs_extra::dir::...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
What about mapping this and `ExpectedStderrFile` by filling up `config.expected_stdout.extend(... read lines from file)`?
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index a4d5abd2fec4..d64104af823c 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -4,8 +4,8 @@ //! for a WASM test. Each configuration represents a distinct test run, with its //! own arguments,...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "What about mapping this and `ExpectedStderrFile` by filling up `config.expected_stdout.extend(... read lines from file)`?", "path": "lib/wasix/tests/wasm_tests/mod.rs", "hunk": "@@ -298,6 +336,19 @@ fn process_directive(\n \"ExpectedStdout\" => {\n config.expected_stdou...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
Here it's going to simplify as well (no extra case for stdout_files and stderr_files.
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index a4d5abd2fec4..d64104af823c 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -4,8 +4,8 @@ //! for a WASM test. Each configuration represents a distinct test run, with its //! own arguments,...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "Here it's going to simplify as well (no extra case for stdout_files and stderr_files.", "path": "lib/wasix/tests/wasm_tests/mod.rs", "hunk": "@@ -516,17 +683,31 @@ fn run_integration_test(config: Config) -> Result<libtest_mimic::Completion> {\n }\n \n if !config.expected_stdout.is_...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
I would not introduce 2 new functions given the previous suggestion.
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index a4d5abd2fec4..d64104af823c 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -4,8 +4,8 @@ //! for a WASM test. Each configuration represents a distinct test run, with its //! own arguments,...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "I would not introduce 2 new functions given the previous suggestion.", "path": "lib/wasix/tests/wasm_tests/mod.rs", "hunk": "@@ -546,6 +727,48 @@ fn run_integration_test(config: Config) -> Result<libtest_mimic::Completion> {\n Ok(libtest_mimic::Completion::Completed)\n }\n \n+fn assert...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
I am thinking if `Self::RustSourceFile`, `Self::CSourceFile` and `Self::CppSourceFile` would not be a better option?
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index a4d5abd2fec4..d64104af823c 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -4,8 +4,8 @@ //! for a WASM test. Each configuration represents a distinct test run, with its //! own arguments,...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "I am thinking if `Self::RustSourceFile`, `Self::CSourceFile` and `Self::CppSourceFile` would not be a better option?", "path": "lib/wasix/tests/wasm_tests/mod.rs", "hunk": "@@ -580,10 +810,14 @@ impl PrimarySource {\n \n fn is_default(&self) -> bool {\n match self {\n- ...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
No need for documenting it.
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/README.md b/lib/wasix/tests/wasm_tests/README.md index d50561813185..0a0d19bfd262 100644 --- a/lib/wasix/tests/wasm_tests/README.md +++ b/lib/wasix/tests/wasm_tests/README.md @@ -18,11 +18,14 @@ The harness builds each discovered fixture, runs the resulting `main` module thr and...
[ "lib/wasix/tests/wasm_tests/README.md" ]
[ { "comment": "No need for documenting it.", "path": "lib/wasix/tests/wasm_tests/README.md", "hunk": "@@ -13,16 +13,21 @@ as a test fixture:\n - `main.c`\n - `main.cpp`\n - `$name.sh`\n+- one or more `*.rs` files\n \n The harness builds each discovered fixture, runs the resulting `main` module through th...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
I would rather mention that a `.rs` test requires the cross `rustc` compiler installed.
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/README.md b/lib/wasix/tests/wasm_tests/README.md index d50561813185..0a0d19bfd262 100644 --- a/lib/wasix/tests/wasm_tests/README.md +++ b/lib/wasix/tests/wasm_tests/README.md @@ -18,11 +18,14 @@ The harness builds each discovered fixture, runs the resulting `main` module thr and...
[ "lib/wasix/tests/wasm_tests/README.md" ]
[ { "comment": "I would rather mention that a `.rs` test requires the cross `rustc` compiler installed.", "path": "lib/wasix/tests/wasm_tests/README.md", "hunk": "@@ -46,3 +52,4 @@ the primary source, for example `//#BuildEnv: WASIXCC_PIC=1` in C/C++ sources or\n These tests run through the normal `wasix`...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
Let's extend the current function - no need for a wrapper.
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/runner.rs b/lib/wasix/tests/wasm_tests/runner.rs index 7b7fbf797ae7..73bd1f77a282 100644 --- a/lib/wasix/tests/wasm_tests/runner.rs +++ b/lib/wasix/tests/wasm_tests/runner.rs @@ -306,6 +306,8 @@ pub(crate) fn run_wasm_with_runner_config( wasm_path: &PathBuf, dir: &Path,...
[ "lib/wasix/tests/wasm_tests/runner.rs" ]
[ { "comment": "Let's extend the current function - no need for a wrapper.", "path": "lib/wasix/tests/wasm_tests/runner.rs", "hunk": "@@ -302,17 +302,38 @@ pub(crate) struct WasmRunResult {\n pub(crate) error: Option<String>,\n }\n \n-pub(crate) fn run_wasm_with_runner_config(\n+pub(crate) fn run_wasm...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
Can we please make it a new "AbstractConfigFile: wasi-fyi.config" directive instead (and explicitly listing it in all the FYI tests files)?
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index a4d5abd2fec4..d64104af823c 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -4,8 +4,8 @@ //! for a WASM test. Each configuration represents a distinct test run, with its //! own arguments,...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "Can we please make it a new \"AbstractConfigFile: wasi-fyi.config\" directive instead (and explicitly listing it in all the FYI tests files)?", "path": "lib/wasix/tests/wasm_tests/mod.rs", "hunk": "@@ -55,9 +72,12 @@ use std::str::FromStr;\n use anyhow::bail;\n use libtest_mimic::Trial;\n ...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
As mentioned, let's make it an explicit directive.
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index a4d5abd2fec4..d64104af823c 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -4,8 +4,8 @@ //! for a WASM test. Each configuration represents a distinct test run, with its //! own arguments,...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "As mentioned, let's make it an explicit directive.", "path": "lib/wasix/tests/wasm_tests/mod.rs", "hunk": "@@ -207,13 +243,41 @@ fn parse_configs(default_config: &Config) -> Result<Vec<Config>> {\n let mut config = default_config.clone();\n let mut build_env = Vec::new();\n \n+ ...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
Let's directly read the file content at this place and put the lines to the `config.expected_stderr` which should remain `Vec<String>`.
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index a4d5abd2fec4..d64104af823c 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -4,8 +4,8 @@ //! for a WASM test. Each configuration represents a distinct test run, with its //! own arguments,...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "Let's directly read the file content at this place and put the lines to the `config.expected_stderr` which should remain `Vec<String>`.", "path": "lib/wasix/tests/wasm_tests/mod.rs", "hunk": "@@ -296,7 +360,36 @@ fn process_directive(\n .collect();\n }\n \"E...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
I would remove `is_rust` and make it based on the previous match statement where we handle `RustSourceFile`.
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index a4d5abd2fec4..d64104af823c 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -4,8 +4,8 @@ //! for a WASM test. Each configuration represents a distinct test run, with its //! own arguments,...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "I would remove `is_rust` and make it based on the previous match statement where we handle `RustSourceFile`.", "path": "lib/wasix/tests/wasm_tests/mod.rs", "hunk": "@@ -404,26 +528,39 @@ fn run_build_script(config: &Config) -> anyhow::Result<PathBuf> {\n .env(\"WASIXCC_DISC...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
I’m curious whether there’s an existing Rust crate that already provides this functionality. To be honest, I’m not a big fan of the custom homegrown approach.
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index a4d5abd2fec4..d64104af823c 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -4,8 +4,8 @@ //! for a WASM test. Each configuration represents a distinct test run, with its //! own arguments,...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "I’m curious whether there’s an existing Rust crate that already provides this functionality. To be honest, I’m not a big fan of the custom homegrown approach.", "path": "lib/wasix/tests/wasm_tests/mod.rs", "hunk": "@@ -442,6 +579,53 @@ fn run_build_script(config: &Config) -> anyhow::Result...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
If the directive is missing, I would not overwrite the program name.
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index a4d5abd2fec4..d64104af823c 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -4,8 +4,8 @@ //! for a WASM test. Each configuration represents a distinct test run, with its //! own arguments,...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "If the directive is missing, I would not overwrite the program name.", "path": "lib/wasix/tests/wasm_tests/mod.rs", "hunk": "@@ -463,40 +647,71 @@ fn run_integration_test(config: Config) -> Result<libtest_mimic::Completion> {\n File::create(run_dir.join(path))?.write_all(file_conte...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
As mentioned, please read it at the time of the Directive parsing.
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index a4d5abd2fec4..d64104af823c 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -4,8 +4,8 @@ //! for a WASM test. Each configuration represents a distinct test run, with its //! own arguments,...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "As mentioned, please read it at the time of the Directive parsing.", "path": "lib/wasix/tests/wasm_tests/mod.rs", "hunk": "@@ -463,40 +647,71 @@ fn run_integration_test(config: Config) -> Result<libtest_mimic::Completion> {\n File::create(run_dir.join(path))?.write_all(file_content...
true
wasmerio/wasmer
6,625
comment_to_fix
test(wasix): port wasi-fyi suite to wasm harness
Please add a comment that explains we allow multiple Rust source files in a same directory.
39bdf3646de17dca8aa428904c6b7754d5147a44
3b2e9d1dfc2b204cbe35b7f73edb4beab82f6e56
diff --git a/lib/wasix/tests/wasm_tests/mod.rs b/lib/wasix/tests/wasm_tests/mod.rs index a4d5abd2fec4..d64104af823c 100644 --- a/lib/wasix/tests/wasm_tests/mod.rs +++ b/lib/wasix/tests/wasm_tests/mod.rs @@ -4,8 +4,8 @@ //! for a WASM test. Each configuration represents a distinct test run, with its //! own arguments,...
[ "lib/wasix/tests/wasm_tests/mod.rs" ]
[ { "comment": "Please add a comment that explains we allow multiple Rust source files in a same directory.", "path": "lib/wasix/tests/wasm_tests/mod.rs", "hunk": "@@ -607,17 +860,53 @@ fn identify_primary_sources(test_src_dir: &Path) -> Result<Vec<PrimarySource>> {\n for file in [\"main.c\", \"main.c...
true
wasmerio/wasmer
6,604
issue_to_patch
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
# Description ## Motivation When targeting platforms without hardware floating-point (e.g. soft-float RISC-V), LLVM emits references to compiler-rt / libgcc arithmetic helpers such as `__adddf3`, `__mulsf3`, `__eqdf2`, etc. Wasmer's LLVM backend resolves libcall symbols by name at JIT link time, but thes...
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/compiler-llvm/src/object_file.rs b/lib/compiler-llvm/src/object_file.rs index 7265bca86cd6..60d8fdf5b7f3 100644 --- a/lib/compiler-llvm/src/object_file.rs +++ b/lib/compiler-llvm/src/object_file.rs @@ -92,6 +92,59 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! { "wasmer_vm...
[ "lib/compiler-llvm/src/object_file.rs", "lib/types/src/libcalls.rs", "lib/vm/src/libcalls.rs" ]
[ { "comment": "The Mach-O libcalls map is keyed by the symbol names that appear as undefined references in LLVM-generated object files (see how `LIBCALLS_MACHO` is consumed in `load_object_file` around line 419). When LLVM emits code that needs `__mulsi3`, the Mach-O object file references the symbol `___mulsi3`...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
The Mach-O libcalls map is keyed by the symbol names that appear as undefined references in LLVM-generated object files (see how `LIBCALLS_MACHO` is consumed in `load_object_file` around line 419). When LLVM emits code that needs `__mulsi3`, the Mach-O object file references the symbol `___mulsi3` (the C name with the ...
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/compiler-llvm/src/object_file.rs b/lib/compiler-llvm/src/object_file.rs index 7265bca86cd6..60d8fdf5b7f3 100644 --- a/lib/compiler-llvm/src/object_file.rs +++ b/lib/compiler-llvm/src/object_file.rs @@ -92,6 +92,59 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! { "wasmer_vm...
[ "lib/compiler-llvm/src/object_file.rs" ]
[ { "comment": "The Mach-O libcalls map is keyed by the symbol names that appear as undefined references in LLVM-generated object files (see how `LIBCALLS_MACHO` is consumed in `load_object_file` around line 419). When LLVM emits code that needs `__mulsi3`, the Mach-O object file references the symbol `___mulsi3`...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
I would like to see it guarded by the `cfg` so that it gets compiled in only for the desired platforms.
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/compiler-llvm/src/object_file.rs b/lib/compiler-llvm/src/object_file.rs index 7265bca86cd6..60d8fdf5b7f3 100644 --- a/lib/compiler-llvm/src/object_file.rs +++ b/lib/compiler-llvm/src/object_file.rs @@ -92,6 +92,59 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! { "wasmer_vm...
[ "lib/compiler-llvm/src/object_file.rs" ]
[ { "comment": "I would like to see it guarded by the `cfg` so that it gets compiled in only for the desired platforms.", "path": "lib/compiler-llvm/src/object_file.rs", "hunk": "@@ -90,6 +90,62 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! {\n \"wasmer_eh_personality2\" => L...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
Do you really need `memset`? It's a risky operation from the VM perspective - can we map it to `wasmer_vm_memory32_fill`?
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/compiler-llvm/src/object_file.rs b/lib/compiler-llvm/src/object_file.rs index 7265bca86cd6..60d8fdf5b7f3 100644 --- a/lib/compiler-llvm/src/object_file.rs +++ b/lib/compiler-llvm/src/object_file.rs @@ -92,6 +92,59 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! { "wasmer_vm...
[ "lib/compiler-llvm/src/object_file.rs" ]
[ { "comment": "Do you really need `memset`? It's a risky operation from the VM perspective - can we map it to `wasmer_vm_memory32_fill`?", "path": "lib/compiler-llvm/src/object_file.rs", "hunk": "@@ -90,6 +90,62 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! {\n \"wasmer_eh_p...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
Is there any supported Mach-O platform that needs these math routines?
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/compiler-llvm/src/object_file.rs b/lib/compiler-llvm/src/object_file.rs index 7265bca86cd6..60d8fdf5b7f3 100644 --- a/lib/compiler-llvm/src/object_file.rs +++ b/lib/compiler-llvm/src/object_file.rs @@ -92,6 +92,59 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! { "wasmer_vm...
[ "lib/compiler-llvm/src/object_file.rs" ]
[ { "comment": "Is there any supported Mach-O platform that needs these math routines?", "path": "lib/compiler-llvm/src/object_file.rs", "hunk": "@@ -90,6 +90,62 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! {\n \"wasmer_eh_personality2\" => LibCall::EHPersonality2,\n \"w...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
Am I right we will hit a linker error if some of the defined routines is undefined?
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/vm/src/libcalls.rs b/lib/vm/src/libcalls.rs index 424e05284a69..81ac7310e3e7 100644 --- a/lib/vm/src/libcalls.rs +++ b/lib/vm/src/libcalls.rs @@ -1006,5 +1006,149 @@ pub fn function_pointer(libcall: LibCall) -> usize { } LibCall::DebugUsize => wasmer_vm_dbg_usize as *const () as usize...
[ "lib/vm/src/libcalls.rs" ]
[ { "comment": "Am I right we will hit a linker error if some of the defined routines is undefined?", "path": "lib/vm/src/libcalls.rs", "hunk": "@@ -1006,5 +1006,127 @@ pub fn function_pointer(libcall: LibCall) -> usize {\n }\n LibCall::DebugUsize => wasmer_vm_dbg_usize as *const () as usi...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
The soft-float map is ELF-specific (unprefixed symbols) but `lookup_libcall` consults it regardless of `fmt`. On Mach-O, these libcalls typically appear with a leading underscore (e.g. `___adddf3`), so they won’t be recognized. Fix by adding a Mach-O soft-float map (underscore-prefixed keys) and selecting the correct o...
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/compiler-llvm/src/object_file.rs b/lib/compiler-llvm/src/object_file.rs index 7265bca86cd6..60d8fdf5b7f3 100644 --- a/lib/compiler-llvm/src/object_file.rs +++ b/lib/compiler-llvm/src/object_file.rs @@ -92,6 +92,59 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! { "wasmer_vm...
[ "lib/compiler-llvm/src/object_file.rs" ]
[ { "comment": "The soft-float map is ELF-specific (unprefixed symbols) but `lookup_libcall` consults it regardless of `fmt`. On Mach-O, these libcalls typically appear with a leading underscore (e.g. `___adddf3`), so they won’t be recognized. Fix by adding a Mach-O soft-float map (underscore-prefixed keys) and s...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
The soft-float map is ELF-specific (unprefixed symbols) but `lookup_libcall` consults it regardless of `fmt`. On Mach-O, these libcalls typically appear with a leading underscore (e.g. `___adddf3`), so they won’t be recognized. Fix by adding a Mach-O soft-float map (underscore-prefixed keys) and selecting the correct o...
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/compiler-llvm/src/object_file.rs b/lib/compiler-llvm/src/object_file.rs index 7265bca86cd6..60d8fdf5b7f3 100644 --- a/lib/compiler-llvm/src/object_file.rs +++ b/lib/compiler-llvm/src/object_file.rs @@ -92,6 +92,59 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! { "wasmer_vm...
[ "lib/compiler-llvm/src/object_file.rs" ]
[ { "comment": "The soft-float map is ELF-specific (unprefixed symbols) but `lookup_libcall` consults it regardless of `fmt`. On Mach-O, these libcalls typically appear with a leading underscore (e.g. `___adddf3`), so they won’t be recognized. Fix by adding a Mach-O soft-float map (underscore-prefixed keys) and s...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
All of these externs are declared with empty signatures, which is misleading and makes it easy for a future change to accidentally call them with the wrong ABI/types. Since their signatures are well-known (e.g., `__adddf3(f64, f64) -> f64`, `__divdi3(i64, i64) -> i64`, etc.), declare them with correct parameter/return ...
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/vm/src/libcalls.rs b/lib/vm/src/libcalls.rs index 424e05284a69..81ac7310e3e7 100644 --- a/lib/vm/src/libcalls.rs +++ b/lib/vm/src/libcalls.rs @@ -1006,5 +1006,149 @@ pub fn function_pointer(libcall: LibCall) -> usize { } LibCall::DebugUsize => wasmer_vm_dbg_usize as *const () as usize...
[ "lib/vm/src/libcalls.rs" ]
[ { "comment": "All of these externs are declared with empty signatures, which is misleading and makes it easy for a future change to accidentally call them with the wrong ABI/types. Since their signatures are well-known (e.g., `__adddf3(f64, f64) -> f64`, `__divdi3(i64, i64) -> i64`, etc.), declare them with cor...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
The exported symbol `wasmer_vm_sqrt` is inconsistent with nearby conventions like `wasmer_vm_f64_floor` / `wasmer_vm_f64_nearest`. Renaming to something like `wasmer_vm_f64_sqrt` (and updating `LibCall::to_symbol` accordingly) would make the API clearer and avoid ambiguity if an `f32` sqrt libcall is added later.
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/vm/src/libcalls.rs b/lib/vm/src/libcalls.rs index 424e05284a69..81ac7310e3e7 100644 --- a/lib/vm/src/libcalls.rs +++ b/lib/vm/src/libcalls.rs @@ -1006,5 +1006,149 @@ pub fn function_pointer(libcall: LibCall) -> usize { } LibCall::DebugUsize => wasmer_vm_dbg_usize as *const () as usize...
[ "lib/vm/src/libcalls.rs" ]
[ { "comment": "The exported symbol `wasmer_vm_sqrt` is inconsistent with nearby conventions like `wasmer_vm_f64_floor` / `wasmer_vm_f64_nearest`. Renaming to something like `wasmer_vm_f64_sqrt` (and updating `LibCall::to_symbol` accordingly) would make the API clearer and avoid ambiguity if an `f32` sqrt libcall...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
These match arms reference many compiler-rt/libgcc symbols unconditionally. Even if the LLVM object loader only ever produces these `LibCall` variants on certain architectures, the Rust binary may still need to resolve the referenced extern symbols at link time. To avoid potential link failures on targets/toolchains th...
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/vm/src/libcalls.rs b/lib/vm/src/libcalls.rs index 424e05284a69..81ac7310e3e7 100644 --- a/lib/vm/src/libcalls.rs +++ b/lib/vm/src/libcalls.rs @@ -1006,5 +1006,149 @@ pub fn function_pointer(libcall: LibCall) -> usize { } LibCall::DebugUsize => wasmer_vm_dbg_usize as *const () as usize...
[ "lib/vm/src/libcalls.rs" ]
[ { "comment": "These match arms reference many compiler-rt/libgcc symbols unconditionally. Even if the LLVM object loader only ever produces these `LibCall` variants on certain architectures, the Rust binary may still need to resolve the referenced extern symbols at link time. To avoid potential link failures on...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
These match arms reference many compiler-rt/libgcc symbols unconditionally. Even if the LLVM object loader only ever produces these `LibCall` variants on certain architectures, the Rust binary may still need to resolve the referenced extern symbols at link time. To avoid potential link failures on targets/toolchains th...
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/vm/src/libcalls.rs b/lib/vm/src/libcalls.rs index 424e05284a69..81ac7310e3e7 100644 --- a/lib/vm/src/libcalls.rs +++ b/lib/vm/src/libcalls.rs @@ -1006,5 +1006,149 @@ pub fn function_pointer(libcall: LibCall) -> usize { } LibCall::DebugUsize => wasmer_vm_dbg_usize as *const () as usize...
[ "lib/vm/src/libcalls.rs" ]
[ { "comment": "These match arms reference many compiler-rt/libgcc symbols unconditionally. Even if the LLVM object loader only ever produces these `LibCall` variants on certain architectures, the Rust binary may still need to resolve the referenced extern symbols at link time. To avoid potential link failures on...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
These match arms take addresses of `__*` soft-float symbols unconditionally, which forces the final Wasmer VM binary to link against them on all targets/builds. This can produce link-time failures on platforms/toolchains where these helpers are not available under these names (or are not linked in by default). Consider...
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/vm/src/libcalls.rs b/lib/vm/src/libcalls.rs index 424e05284a69..81ac7310e3e7 100644 --- a/lib/vm/src/libcalls.rs +++ b/lib/vm/src/libcalls.rs @@ -1006,5 +1006,149 @@ pub fn function_pointer(libcall: LibCall) -> usize { } LibCall::DebugUsize => wasmer_vm_dbg_usize as *const () as usize...
[ "lib/vm/src/libcalls.rs" ]
[ { "comment": "These match arms take addresses of `__*` soft-float symbols unconditionally, which forces the final Wasmer VM binary to link against them on all targets/builds. This can produce link-time failures on platforms/toolchains where these helpers are not available under these names (or are not linked in...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
Using `unsafe extern \"C\" { ... }` for the extern block can be MSRV-sensitive depending on which Rust version this repo supports. If MSRV compatibility is a goal, prefer the broadly-supported form `extern \"C\" { ... }` (calls remain `unsafe` at the call site).
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/vm/src/libcalls.rs b/lib/vm/src/libcalls.rs index 424e05284a69..81ac7310e3e7 100644 --- a/lib/vm/src/libcalls.rs +++ b/lib/vm/src/libcalls.rs @@ -1006,5 +1006,149 @@ pub fn function_pointer(libcall: LibCall) -> usize { } LibCall::DebugUsize => wasmer_vm_dbg_usize as *const () as usize...
[ "lib/vm/src/libcalls.rs" ]
[ { "comment": "Using `unsafe extern \\\"C\\\" { ... }` for the extern block can be MSRV-sensitive depending on which Rust version this repo supports. If MSRV compatibility is a goal, prefer the broadly-supported form `extern \\\"C\\\" { ... }` (calls remain `unsafe` at the call site).\n", "path": "lib/vm/src...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
Keying soft-float libcall recognition off the build-time `target_arch` risks failing to resolve relocations if an object file contains `__addsf3`-style symbols on x86_64/aarch64 (e.g., unusual target settings / soft-float configurations / cross-target object parsing). Since this is just a name->enum lookup, consider ma...
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/compiler-llvm/src/object_file.rs b/lib/compiler-llvm/src/object_file.rs index 7265bca86cd6..60d8fdf5b7f3 100644 --- a/lib/compiler-llvm/src/object_file.rs +++ b/lib/compiler-llvm/src/object_file.rs @@ -92,6 +92,59 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! { "wasmer_vm...
[ "lib/compiler-llvm/src/object_file.rs" ]
[ { "comment": "Keying soft-float libcall recognition off the build-time `target_arch` risks failing to resolve relocations if an object file contains `__addsf3`-style symbols on x86_64/aarch64 (e.g., unusual target settings / soft-float configurations / cross-target object parsing). Since this is just a name->en...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
This changes behavior from ignoring `Err` results (`if let Ok(Some(...))`) to propagating them. If `symbol_name_to_relocation_target` can return `Err` for non-fatal/\"not a match\" cases, this will turn previously-successful compilations into hard failures. A safer pattern is to ensure `symbol_name_to_relocation_target...
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/compiler-llvm/src/object_file.rs b/lib/compiler-llvm/src/object_file.rs index 7265bca86cd6..60d8fdf5b7f3 100644 --- a/lib/compiler-llvm/src/object_file.rs +++ b/lib/compiler-llvm/src/object_file.rs @@ -92,6 +92,59 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! { "wasmer_vm...
[ "lib/compiler-llvm/src/object_file.rs" ]
[ { "comment": "This changes behavior from ignoring `Err` results (`if let Ok(Some(...))`) to propagating them. If `symbol_name_to_relocation_target` can return `Err` for non-fatal/\\\"not a match\\\" cases, this will turn previously-successful compilations into hard failures. A safer pattern is to ensure `symbol...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
Can we have a rather explicit listing of architectures where we want the soft FP operations? `!x86_64 || !aarch64` does not seem correct to me.
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/compiler-llvm/src/object_file.rs b/lib/compiler-llvm/src/object_file.rs index 7265bca86cd6..60d8fdf5b7f3 100644 --- a/lib/compiler-llvm/src/object_file.rs +++ b/lib/compiler-llvm/src/object_file.rs @@ -92,6 +92,59 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! { "wasmer_vm...
[ "lib/compiler-llvm/src/object_file.rs" ]
[ { "comment": "Can we have a rather explicit listing of architectures where we want the soft FP operations? `!x86_64 || !aarch64` does not seem correct to me.", "path": "lib/compiler-llvm/src/object_file.rs", "hunk": "@@ -92,6 +92,58 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map!...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
Similarly here.
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/compiler-llvm/src/object_file.rs b/lib/compiler-llvm/src/object_file.rs index 7265bca86cd6..60d8fdf5b7f3 100644 --- a/lib/compiler-llvm/src/object_file.rs +++ b/lib/compiler-llvm/src/object_file.rs @@ -92,6 +92,59 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! { "wasmer_vm...
[ "lib/compiler-llvm/src/object_file.rs" ]
[ { "comment": "Similarly here.", "path": "lib/compiler-llvm/src/object_file.rs", "hunk": "@@ -152,6 +204,24 @@ static LIBCALLS_MACHO: phf::Map<&'static str, LibCall> = phf::phf_map! {\n \"_wasmer_vm_dbg_str\" => LibCall::DebugStr,\n };\n \n+fn lookup_libcall(name: &str, fmt: BinaryFormat) -> Option<L...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
Can we just one `#[cfg]` gate where we'll do a sub-match for all the soft FP calls?
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/vm/src/libcalls.rs b/lib/vm/src/libcalls.rs index 424e05284a69..81ac7310e3e7 100644 --- a/lib/vm/src/libcalls.rs +++ b/lib/vm/src/libcalls.rs @@ -1006,5 +1006,149 @@ pub fn function_pointer(libcall: LibCall) -> usize { } LibCall::DebugUsize => wasmer_vm_dbg_usize as *const () as usize...
[ "lib/vm/src/libcalls.rs" ]
[ { "comment": "Can we just one `#[cfg]` gate where we'll do a sub-match for all the soft FP calls?", "path": "lib/vm/src/libcalls.rs", "hunk": "@@ -1006,5 +1006,189 @@ pub fn function_pointer(libcall: LibCall) -> usize {\n }\n LibCall::DebugUsize => wasmer_vm_dbg_usize as *const () as usi...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
I think it's still imprecise - we should combine the target architecture with `"target_feature = "f"`?
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/compiler-llvm/src/object_file.rs b/lib/compiler-llvm/src/object_file.rs index 7265bca86cd6..60d8fdf5b7f3 100644 --- a/lib/compiler-llvm/src/object_file.rs +++ b/lib/compiler-llvm/src/object_file.rs @@ -92,6 +92,59 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! { "wasmer_vm...
[ "lib/compiler-llvm/src/object_file.rs" ]
[ { "comment": "I think it's still imprecise - we should combine the target architecture with `\"target_feature = \"f\"`?", "path": "lib/compiler-llvm/src/object_file.rs", "hunk": "@@ -92,10 +92,8 @@ static LIBCALLS_ELF: phf::Map<&'static str, LibCall> = phf::phf_map! {\n \"wasmer_vm_dbg_str\" => LibC...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
Should be aligned with the previously used `cfg` selection.
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/vm/src/libcalls.rs b/lib/vm/src/libcalls.rs index 424e05284a69..81ac7310e3e7 100644 --- a/lib/vm/src/libcalls.rs +++ b/lib/vm/src/libcalls.rs @@ -1006,5 +1006,149 @@ pub fn function_pointer(libcall: LibCall) -> usize { } LibCall::DebugUsize => wasmer_vm_dbg_usize as *const () as usize...
[ "lib/vm/src/libcalls.rs" ]
[ { "comment": "Should be aligned with the previously used `cfg` selection.", "path": "lib/vm/src/libcalls.rs", "hunk": "@@ -1006,5 +1006,149 @@ pub fn function_pointer(libcall: LibCall) -> usize {\n }\n LibCall::DebugUsize => wasmer_vm_dbg_usize as *const () as usize,\n LibCall::D...
true
wasmerio/wasmer
6,604
comment_to_fix
Add soft-float libcall support for LLVM backend on hardware-FP-less targets
Same here.
ee17dece6c8b56fc920a90ef345264c06292f70a
71573bf0ecf54dd4db2953727793f49d8a384c4d
diff --git a/lib/vm/src/libcalls.rs b/lib/vm/src/libcalls.rs index 424e05284a69..81ac7310e3e7 100644 --- a/lib/vm/src/libcalls.rs +++ b/lib/vm/src/libcalls.rs @@ -1006,5 +1006,149 @@ pub fn function_pointer(libcall: LibCall) -> usize { } LibCall::DebugUsize => wasmer_vm_dbg_usize as *const () as usize...
[ "lib/vm/src/libcalls.rs" ]
[ { "comment": "Same here.", "path": "lib/vm/src/libcalls.rs", "hunk": "@@ -1006,5 +1006,149 @@ pub fn function_pointer(libcall: LibCall) -> usize {\n }\n LibCall::DebugUsize => wasmer_vm_dbg_usize as *const () as usize,\n LibCall::DebugStr => wasmer_vm_dbg_str as *const () as usiz...
true
wasmerio/wasmer
6,603
issue_to_patch
Add `Artifact::finished_function_extents()` to expose compiled function body locations
# Description Adds `Artifact::finished_function_extents()` and exposes access to the underlying `Artifact` via `Module::sys_artifact()`, making it possible to retrieve the in-memory location and byte length of each locally-defined function body in a compiled module. ## Motivation When a compiled module is exec...
4263a29798a884562904cb8a9e8643358957bfae
6e1c967952fa549a1d16c56a2995c202ff2b15d5
diff --git a/lib/api/src/backend/sys/entities/module.rs b/lib/api/src/backend/sys/entities/module.rs index 4e6ae276fa1d..2ce2b9dc4e02 100644 --- a/lib/api/src/backend/sys/entities/module.rs +++ b/lib/api/src/backend/sys/entities/module.rs @@ -287,4 +287,21 @@ impl crate::Module { _ => panic!("Not a `sys` m...
[ "lib/api/src/backend/sys/entities/module.rs", "lib/api/src/backend/sys/mod.rs", "lib/api/tests/module.rs", "lib/compiler/src/engine/artifact.rs" ]
[ { "comment": "This calls `Artifact::finished_function_extents()`, but `wasmer_compiler::Artifact` does not define that method in the current codebase. The crate will fail to compile until an accessor for the stored function pointers and lengths is added or this code is changed to use existing APIs.\n", "pat...
diff --git a/lib/api/tests/module.rs b/lib/api/tests/module.rs index df286eb3ee78..7824eb070015 100644 --- a/lib/api/tests/module.rs +++ b/lib/api/tests/module.rs @@ -296,6 +296,106 @@ fn module_custom_sections() -> Result<(), String> { Ok(()) } +#[cfg(all( + not(target_arch = "wasm32"), + any(feature = "...
true
wasmerio/wasmer
6,603
comment_to_fix
Add `Artifact::finished_function_extents()` to expose compiled function body locations
This calls `Artifact::finished_function_extents()`, but `wasmer_compiler::Artifact` does not define that method in the current codebase. The crate will fail to compile until an accessor for the stored function pointers and lengths is added or this code is changed to use existing APIs.
4263a29798a884562904cb8a9e8643358957bfae
6e1c967952fa549a1d16c56a2995c202ff2b15d5
diff --git a/lib/api/src/backend/sys/entities/module.rs b/lib/api/src/backend/sys/entities/module.rs index 4e6ae276fa1d..2ce2b9dc4e02 100644 --- a/lib/api/src/backend/sys/entities/module.rs +++ b/lib/api/src/backend/sys/entities/module.rs @@ -287,4 +287,21 @@ impl crate::Module { _ => panic!("Not a `sys` m...
[ "lib/api/src/backend/sys/entities/module.rs" ]
[ { "comment": "This calls `Artifact::finished_function_extents()`, but `wasmer_compiler::Artifact` does not define that method in the current codebase. The crate will fail to compile until an accessor for the stored function pointers and lengths is added or this code is changed to use existing APIs.\n", "pat...
true
wasmerio/wasmer
6,603
comment_to_fix
Add `Artifact::finished_function_extents()` to expose compiled function body locations
These tests are only gated on the target architecture, but the new API explicitly returns an empty vector for non-`sys` backends. In native builds where the default engine is `v8`/`js` or `sys` is not enabled, this assertion expects three entries and will fail; restrict the positive assertions to the `sys` backend or c...
4263a29798a884562904cb8a9e8643358957bfae
6e1c967952fa549a1d16c56a2995c202ff2b15d5
diff --git a/lib/api/tests/module.rs b/lib/api/tests/module.rs index df286eb3ee78..7824eb070015 100644 --- a/lib/api/tests/module.rs +++ b/lib/api/tests/module.rs @@ -296,6 +296,106 @@ fn module_custom_sections() -> Result<(), String> { Ok(()) } +#[cfg(all( + not(target_arch = "wasm32"), + any(feature = "...
[ "lib/api/tests/module.rs" ]
[ { "comment": "These tests are only gated on the target architecture, but the new API explicitly returns an empty vector for non-`sys` backends. In native builds where the default engine is `v8`/`js` or `sys` is not enabled, this assertion expects three entries and will fail; restrict the positive assertions to ...
true
wasmerio/wasmer
6,603
comment_to_fix
Add `Artifact::finished_function_extents()` to expose compiled function body locations
This imported-function test only checks the number of extents, so an implementation that returns global function indices like `[1, 2]` instead of the promised local indices `[0, 1]` would still pass. Add an assertion for the returned indices here to cover the imported-function offset behavior described by the API.
4263a29798a884562904cb8a9e8643358957bfae
6e1c967952fa549a1d16c56a2995c202ff2b15d5
diff --git a/lib/api/tests/module.rs b/lib/api/tests/module.rs index df286eb3ee78..7824eb070015 100644 --- a/lib/api/tests/module.rs +++ b/lib/api/tests/module.rs @@ -296,6 +296,106 @@ fn module_custom_sections() -> Result<(), String> { Ok(()) } +#[cfg(all( + not(target_arch = "wasm32"), + any(feature = "...
[ "lib/api/tests/module.rs" ]
[ { "comment": "This imported-function test only checks the number of extents, so an implementation that returns global function indices like `[1, 2]` instead of the promised local indices `[0, 1]` would still pass. Add an assertion for the returned indices here to cover the imported-function offset behavior desc...
true
wasmerio/wasmer
6,603
comment_to_fix
Add `Artifact::finished_function_extents()` to expose compiled function body locations
This will expose invalid zero-length extents for modules loaded via `static-artifact-load`: `Artifact::deserialize_object` initializes every `finished_function_lengths` entry to `0` (lib/compiler/src/engine/artifact.rs:1281-1284). Either preserve real lengths for static artifacts or return an empty result for that unsu...
4263a29798a884562904cb8a9e8643358957bfae
6e1c967952fa549a1d16c56a2995c202ff2b15d5
diff --git a/lib/api/src/backend/sys/entities/module.rs b/lib/api/src/backend/sys/entities/module.rs index 4e6ae276fa1d..2ce2b9dc4e02 100644 --- a/lib/api/src/backend/sys/entities/module.rs +++ b/lib/api/src/backend/sys/entities/module.rs @@ -287,4 +287,21 @@ impl crate::Module { _ => panic!("Not a `sys` m...
[ "lib/api/src/backend/sys/entities/module.rs" ]
[ { "comment": "This will expose invalid zero-length extents for modules loaded via `static-artifact-load`: `Artifact::deserialize_object` initializes every `finished_function_lengths` entry to `0` (lib/compiler/src/engine/artifact.rs:1281-1284). Either preserve real lengths for static artifacts or return an empt...
true
wasmerio/wasmer
6,603
comment_to_fix
Add `Artifact::finished_function_extents()` to expose compiled function body locations
`finished_function_extents()` assumes the artifact has been allocated, but `Artifact::from_parts` explicitly leaves `allocated` as `None` for non-native targets used for cross-compilation. Calling the new public `Module::function_extents()` on such a sys module will therefore panic instead of returning an empty list fo...
4263a29798a884562904cb8a9e8643358957bfae
6e1c967952fa549a1d16c56a2995c202ff2b15d5
diff --git a/lib/compiler/src/engine/artifact.rs b/lib/compiler/src/engine/artifact.rs index 9c03a0ec9956..f2be6afe9d49 100644 --- a/lib/compiler/src/engine/artifact.rs +++ b/lib/compiler/src/engine/artifact.rs @@ -786,6 +786,38 @@ impl Artifact { .finished_functions } + /// Returns the start add...
[ "lib/compiler/src/engine/artifact.rs" ]
[ { "comment": "`finished_function_extents()` assumes the artifact has been allocated, but `Artifact::from_parts` explicitly leaves `allocated` as `None` for non-native targets used for cross-compilation. Calling the new public `Module::function_extents()` on such a sys module will therefore panic instead of retu...
true