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
ClickHouse/ClickHouse
107,346
issue_to_patch
Backport #107071 to 26.5: Support trim character sets longer than 16 in trim* functions
Original pull-request https://github.com/ClickHouse/ClickHouse/pull/107071 Cherry-pick pull-request https://github.com/ClickHouse/ClickHouse/pull/107177 This pull-request is a last step of an automated backporting. Treat it as a standard pull-request: look at the checks and resolve conflicts. Merge it only if you inte...
503b8321beb136b493f594dd79efaa00bcf6c6c4
821c3c723af96cb00ae75ee584786ffd524bbd3c
diff --git a/src/Functions/trim.cpp b/src/Functions/trim.cpp index fbbfeab7635d..ea4c4b35a04c 100644 --- a/src/Functions/trim.cpp +++ b/src/Functions/trim.cpp @@ -7,18 +7,24 @@ #include <Functions/FunctionHelpers.h> #include <base/find_symbols.h> +#include <array> + namespace DB { namespace ErrorCodes { ex...
[ "src/Functions/trim.cpp", "tests/queries/0_stateless/03784_functions_various_bugs.reference", "tests/queries/0_stateless/03784_functions_various_bugs.sql", "tests/queries/0_stateless/04051_functions_various_bugs.reference", "tests/queries/0_stateless/04051_functions_various_bugs.sql", "tests/queries/0_sta...
[]
diff --git a/tests/queries/0_stateless/03784_functions_various_bugs.reference b/tests/queries/0_stateless/03784_functions_various_bugs.reference index 4f0092a0bd13..14f5097d57b8 100644 --- a/tests/queries/0_stateless/03784_functions_various_bugs.reference +++ b/tests/queries/0_stateless/03784_functions_various_bugs.ref...
true
ClickHouse/ClickHouse
107,345
issue_to_patch
Backport #107071 to 26.4: Support trim character sets longer than 16 in trim* functions
Original pull-request https://github.com/ClickHouse/ClickHouse/pull/107071 Cherry-pick pull-request https://github.com/ClickHouse/ClickHouse/pull/107176 This pull-request is a last step of an automated backporting. Treat it as a standard pull-request: look at the checks and resolve conflicts. Merge it only if you inte...
7a7d19151f4f1b72f4d39d6d5f4681a147a86704
89d221e6ba94850bc9f7bbff6246d6d1c155a91e
diff --git a/src/Functions/trim.cpp b/src/Functions/trim.cpp index fbbfeab7635d..ea4c4b35a04c 100644 --- a/src/Functions/trim.cpp +++ b/src/Functions/trim.cpp @@ -7,18 +7,24 @@ #include <Functions/FunctionHelpers.h> #include <base/find_symbols.h> +#include <array> + namespace DB { namespace ErrorCodes { ex...
[ "src/Functions/trim.cpp", "tests/queries/0_stateless/03784_functions_various_bugs.reference", "tests/queries/0_stateless/03784_functions_various_bugs.sql", "tests/queries/0_stateless/04051_functions_various_bugs.reference", "tests/queries/0_stateless/04051_functions_various_bugs.sql", "tests/queries/0_sta...
[]
diff --git a/tests/queries/0_stateless/03784_functions_various_bugs.reference b/tests/queries/0_stateless/03784_functions_various_bugs.reference index 4f0092a0bd13..14f5097d57b8 100644 --- a/tests/queries/0_stateless/03784_functions_various_bugs.reference +++ b/tests/queries/0_stateless/03784_functions_various_bugs.ref...
true
ClickHouse/ClickHouse
107,404
issue_to_patch
Backport #106507 to 26.4: Fix wrong-results equality with NaN-producing key function
Original pull-request https://github.com/ClickHouse/ClickHouse/pull/106507 Cherry-pick pull-request https://github.com/ClickHouse/ClickHouse/pull/107403 This pull-request is a last step of an automated backporting. Treat it as a standard pull-request: look at the checks and resolve conflicts. Merge it only if you inte...
9606dc9fee640e794992e5e3bc76447de78fe727
50d133ad5475fa6e84282b834befad299b5e27bb
diff --git a/src/Storages/MergeTree/KeyCondition.cpp b/src/Storages/MergeTree/KeyCondition.cpp index f072b15417b9..d33a22de871e 100644 --- a/src/Storages/MergeTree/KeyCondition.cpp +++ b/src/Storages/MergeTree/KeyCondition.cpp @@ -1927,7 +1927,14 @@ bool KeyCondition::canConstantBeWrappedByDeterministicFunctions( ...
[ "src/Storages/MergeTree/KeyCondition.cpp", "tests/queries/0_stateless/04306_keycondition_nan_equality.reference", "tests/queries/0_stateless/04306_keycondition_nan_equality.sql" ]
[]
diff --git a/tests/queries/0_stateless/04306_keycondition_nan_equality.reference b/tests/queries/0_stateless/04306_keycondition_nan_equality.reference new file mode 100644 index 000000000000..de3badc3ed7f --- /dev/null +++ b/tests/queries/0_stateless/04306_keycondition_nan_equality.reference @@ -0,0 +1,14 @@ +1 +1 +1 +...
true
ClickHouse/ClickHouse
107,406
issue_to_patch
Backport #106507 to 26.5: Fix wrong-results equality with NaN-producing key function
Original pull-request https://github.com/ClickHouse/ClickHouse/pull/106507 Cherry-pick pull-request https://github.com/ClickHouse/ClickHouse/pull/107405 This pull-request is a last step of an automated backporting. Treat it as a standard pull-request: look at the checks and resolve conflicts. Merge it only if you inte...
503b8321beb136b493f594dd79efaa00bcf6c6c4
fabfda40ff61861206cadbd1e49d82901c35612c
diff --git a/src/Storages/MergeTree/KeyCondition.cpp b/src/Storages/MergeTree/KeyCondition.cpp index 91aa8bc5f6ad..6b4360282121 100644 --- a/src/Storages/MergeTree/KeyCondition.cpp +++ b/src/Storages/MergeTree/KeyCondition.cpp @@ -2097,7 +2097,14 @@ bool KeyCondition::canConstantBeWrappedByDeterministicFunctions( ...
[ "src/Storages/MergeTree/KeyCondition.cpp", "tests/queries/0_stateless/04306_keycondition_nan_equality.reference", "tests/queries/0_stateless/04306_keycondition_nan_equality.sql" ]
[]
diff --git a/tests/queries/0_stateless/04306_keycondition_nan_equality.reference b/tests/queries/0_stateless/04306_keycondition_nan_equality.reference new file mode 100644 index 000000000000..de3badc3ed7f --- /dev/null +++ b/tests/queries/0_stateless/04306_keycondition_nan_equality.reference @@ -0,0 +1,14 @@ +1 +1 +1 +...
true
ClickHouse/ClickHouse
105,555
issue_to_patch
`applyTrivialInsertSelectOptimization` drops `columns_description` from context, breaking table function structure hints _Found via ClickGap automated review. Please close or comment if this is incorrect or needs adjustment._ _Retrospective finding from a historical scan of [PR #91599](https://github.com/ClickHouse/C...
Fix usage of insertion table in table functions when `optimize_trivial_insert_select` is enabled
### Changelog category (leave one): - Bug Fix (user-visible misbehavior in an official stable release) ### Changelog entry (a [user-readable short description](https://github.com/ClickHouse/ClickHouse/blob/master/docs/changelog_entry_guidelines.md) of the changes that goes into CHANGELOG.md): Fix usage of insert...
9b22faf1eca3a6d0f481985027d5d26eadcfad0c
f545887f70a4b870d32bf49cd18e13e85b717081
diff --git a/src/Interpreters/Access/InterpreterExecuteAsQuery.cpp b/src/Interpreters/Access/InterpreterExecuteAsQuery.cpp index 8ab1a453a5ba..cc095e3e0b3e 100644 --- a/src/Interpreters/Access/InterpreterExecuteAsQuery.cpp +++ b/src/Interpreters/Access/InterpreterExecuteAsQuery.cpp @@ -33,7 +33,7 @@ namespace ...
[ "src/Interpreters/Access/InterpreterExecuteAsQuery.cpp", "src/Interpreters/InterpreterInsertQuery.cpp", "src/Storages/StorageInMemoryMetadata.cpp", "tests/queries/0_stateless/04266_trivial_insert_select_table_function_structure_hint.reference", "tests/queries/0_stateless/04266_trivial_insert_select_table_fu...
[]
diff --git a/tests/queries/0_stateless/04266_trivial_insert_select_table_function_structure_hint.reference b/tests/queries/0_stateless/04266_trivial_insert_select_table_function_structure_hint.reference new file mode 100644 index 000000000000..641f5b5bad4f --- /dev/null +++ b/tests/queries/0_stateless/04266_trivial_ins...
true
ClickHouse/ClickHouse
107,384
issue_to_patch
read auth_response length as UInt8 in MySQL handshake
`HandshakeResponse::readPayloadImpl` reads the `auth_response` length for the `CLIENT_SECURE_CONNECTION` path into a signed `char` and widens it with `static_cast<unsigned int>`. The matching writer emits this field as an unsigned byte (`static_cast<UInt8>(auth_response.size())`), so the two sides disagree once the hig...
e78a4dac750cb4e71fb1928c6e3d760817d45e78
4668a438051d5f17a5d7ecbb24eb2d380f8eb210
diff --git a/src/Core/MySQL/PacketsConnection.cpp b/src/Core/MySQL/PacketsConnection.cpp index 0cc97f6a8df2..db127898ab3a 100644 --- a/src/Core/MySQL/PacketsConnection.cpp +++ b/src/Core/MySQL/PacketsConnection.cpp @@ -148,9 +148,9 @@ void HandshakeResponse::readPayloadImpl(ReadBuffer & payload) } else if (ca...
[ "src/Core/MySQL/PacketsConnection.cpp", "tests/queries/0_stateless/04328_mysql_handshake_auth_response_length.py", "tests/queries/0_stateless/04328_mysql_handshake_auth_response_length.reference" ]
[ { "comment": "```suggestion\n readBinary(len, payload);\n```", "path": "src/Core/MySQL/PacketsConnection.cpp", "hunk": "@@ -148,9 +148,9 @@ void HandshakeResponse::readPayloadImpl(ReadBuffer & payload)\n }\n else if (capability_flags & CLIENT_SECURE_CONNECTION)\n {\n- char len ...
diff --git a/tests/queries/0_stateless/04328_mysql_handshake_auth_response_length.py b/tests/queries/0_stateless/04328_mysql_handshake_auth_response_length.py new file mode 100755 index 000000000000..6381721adf28 --- /dev/null +++ b/tests/queries/0_stateless/04328_mysql_handshake_auth_response_length.py @@ -0,0 +1,112 ...
true
ClickHouse/ClickHouse
107,384
comment_to_fix
read auth_response length as UInt8 in MySQL handshake
```suggestion readBinary(len, payload); ```
e78a4dac750cb4e71fb1928c6e3d760817d45e78
4668a438051d5f17a5d7ecbb24eb2d380f8eb210
diff --git a/src/Core/MySQL/PacketsConnection.cpp b/src/Core/MySQL/PacketsConnection.cpp index 0cc97f6a8df2..db127898ab3a 100644 --- a/src/Core/MySQL/PacketsConnection.cpp +++ b/src/Core/MySQL/PacketsConnection.cpp @@ -148,9 +148,9 @@ void HandshakeResponse::readPayloadImpl(ReadBuffer & payload) } else if (ca...
[ "src/Core/MySQL/PacketsConnection.cpp" ]
[ { "comment": "```suggestion\n readBinary(len, payload);\n```", "path": "src/Core/MySQL/PacketsConnection.cpp", "hunk": "@@ -148,9 +148,9 @@ void HandshakeResponse::readPayloadImpl(ReadBuffer & payload)\n }\n else if (capability_flags & CLIENT_SECURE_CONNECTION)\n {\n- char len ...
true
ClickHouse/ClickHouse
101,942
issue_to_patch
OOM Canary ### Company or project name ClickHouse ### Use case `clickhouse-server` should be notified before the OOM killer has a chance to kill it. We can implement it by launching a small child process with an inflated OOM score adjustment. If the process is killed, we know that it was the OOM killer, and we can ...
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
Closes: https://github.com/ClickHouse/ClickHouse/issues/101749 ### Changelog category (leave one): - Experimental Feature ### Changelog entry (a user-readable short description of the changes that goes into CHANGELOG.md): Added an experimental Linux OOM canary: a sacrificial child process that the Linux OOM killer ta...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/programs/CMakeLists.txt b/programs/CMakeLists.txt index c44a71c1a85a..10300277822a 100644 --- a/programs/CMakeLists.txt +++ b/programs/CMakeLists.txt @@ -109,6 +109,7 @@ add_subdirectory (keeper-data-dumper) add_subdirectory (keeper-utils) add_subdirectory (local) add_subdirectory (obfuscator) +add_subd...
[ "programs/CMakeLists.txt", "programs/main.cpp", "programs/oom-canary/CMakeLists.txt", "programs/oom-canary/OomCanary.cpp", "programs/oom-canary/clickhouse-oom-canary.cpp", "programs/server/Server.cpp", "src/CMakeLists.txt", "src/Common/FailPoint.cpp", "src/Common/OOMCanary/OOMCanary.cpp", "src/Com...
[ { "comment": "`vfork` is unsafe here for a long-lived canary process and can block server startup.\n\nWith `vfork`, the parent is suspended until the child calls `execve` or `_exit`. In this implementation the child takes the `pid == 0` branch, enters `childMain`, and then loops forever in `sigsuspend` without ...
diff --git a/tests/integration/test_oom_canary/__init__.py b/tests/integration/test_oom_canary/__init__.py new file mode 100644 index 000000000000..e69de29bb2d1 diff --git a/tests/integration/test_oom_canary/configs/enable_off.xml b/tests/integration/test_oom_canary/configs/enable_off.xml new file mode 100644 index 000...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
`oom_canary_enable` defaults to `true`, so this high-impact behavior is enabled immediately for all deployments: once the canary receives `SIGKILL`, the server executes emergency actions (query cancellation, merge cancellation, crash-log path). For safer rollout, this should be opt-in first (for example default `false`...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Core/ServerSettings.cpp b/src/Core/ServerSettings.cpp index ee0a3d9a0d13..a818c288cec6 100644 --- a/src/Core/ServerSettings.cpp +++ b/src/Core/ServerSettings.cpp @@ -1505,6 +1505,45 @@ The policy on how to perform a scheduling of CPU slots specified by `concurrent_ ``` )", 0) \ DECLARE(I...
[ "src/Core/ServerSettings.cpp" ]
[ { "comment": "`oom_canary_enable` defaults to `true`, so this high-impact behavior is enabled immediately for all deployments: once the canary receives `SIGKILL`, the server executes emergency actions (query cancellation, merge cancellation, crash-log path). For safer rollout, this should be opt-in first (for e...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
This introduces high-impact server behavior (global query cancellation, merge cancellation, and a `system.crash_log` write path) behind a regular setting (`oom_canary_enable`) but without an experimental gate. Per ClickHouse rollout rules for risky new server behavior, it should be guarded by an explicit `allow_experi...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Core/ServerSettings.cpp b/src/Core/ServerSettings.cpp index ee0a3d9a0d13..a818c288cec6 100644 --- a/src/Core/ServerSettings.cpp +++ b/src/Core/ServerSettings.cpp @@ -1505,6 +1505,45 @@ The policy on how to perform a scheduling of CPU slots specified by `concurrent_ ``` )", 0) \ DECLARE(I...
[ "src/Core/ServerSettings.cpp" ]
[ { "comment": "This introduces high-impact server behavior (global query cancellation, merge cancellation, and a `system.crash_log` write path) behind a regular setting (`oom_canary_enable`) but without an experimental gate.\n\nPer ClickHouse rollout rules for risky new server behavior, it should be guarded by a...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
This is experimentally gated behavior (`allow_experimental_oom_canary`) but the PR template still uses `Changelog category: New Feature` and leaves the documentation checkbox unchecked. Per template/conventions, this should be categorized as `Experimental Feature`, and documentation should be provided (or the docs blo...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Core/ServerSettings.cpp b/src/Core/ServerSettings.cpp index ee0a3d9a0d13..a818c288cec6 100644 --- a/src/Core/ServerSettings.cpp +++ b/src/Core/ServerSettings.cpp @@ -1505,6 +1505,45 @@ The policy on how to perform a scheduling of CPU slots specified by `concurrent_ ``` )", 0) \ DECLARE(I...
[ "src/Core/ServerSettings.cpp" ]
[ { "comment": "This is experimentally gated behavior (`allow_experimental_oom_canary`) but the PR template still uses `Changelog category: New Feature` and leaves the documentation checkbox unchecked.\n\nPer template/conventions, this should be categorized as `Experimental Feature`, and documentation should be p...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
If `syscall_pidfd_open` fails, this branch goes straight to blocking `reapChild`. For failures like `ENOSYS` (kernel without `pidfd_open`) or `EMFILE`, the canary process is still alive in `pause`, so the monitor thread blocks forever here and `OOMCanary::stop` can hang on `join` during shutdown. Please terminate the ...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Common/OOMCanary/OOMCanary.cpp b/src/Common/OOMCanary/OOMCanary.cpp new file mode 100644 index 000000000000..69f795a2ba0e --- /dev/null +++ b/src/Common/OOMCanary/OOMCanary.cpp @@ -0,0 +1,476 @@ +#include <Common/OOMCanary/OOMCanary.h> + +#if defined(OS_LINUX) + +#include <Common/Epoll.h> +#include <Co...
[ "src/Common/OOMCanary/OOMCanary.cpp" ]
[ { "comment": "If `syscall_pidfd_open` fails, this branch goes straight to blocking `reapChild`. For failures like `ENOSYS` (kernel without `pidfd_open`) or `EMFILE`, the canary process is still alive in `pause`, so the monitor thread blocks forever here and `OOMCanary::stop` can hang on `join` during shutdown.\...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
`oom-canary` is built unconditionally, but this source is Linux-only (`sys/prctl.h`, `PR_SET_PDEATHSIG`, `oom_score_adj`, `close_range` syscall). On non-Linux builds this will fail at compile time. Please gate this translation unit/target to Linux (or provide a non-Linux stub `mainEntryClickHouseOomCanary`), and keep ...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/programs/oom-canary/OomCanary.cpp b/programs/oom-canary/OomCanary.cpp new file mode 100644 index 000000000000..343410e486d6 --- /dev/null +++ b/programs/oom-canary/OomCanary.cpp @@ -0,0 +1,144 @@ +#include <Common/OOMCanaryExitCodes.h> + +#if defined(OS_LINUX) + +#include <Common/Exception.h> +#include <IO...
[ "programs/oom-canary/OomCanary.cpp" ]
[ { "comment": "`oom-canary` is built unconditionally, but this source is Linux-only (`sys/prctl.h`, `PR_SET_PDEATHSIG`, `oom_score_adj`, `close_range` syscall). On non-Linux builds this will fail at compile time.\n\nPlease gate this translation unit/target to Linux (or provide a non-Linux stub `mainEntryClickHou...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
The OOM response contract here includes more than `system.crash_log`: `onCanaryOOM` also calls `ProcessList::killAllQueries` and `MergeList::cancelAll`. Current coverage only proves relaunch + crash-log insertion, so a regression where query/merge cancellation stops working would still pass. Please add one focused as...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/tests/integration/test_oom_canary/test.py b/tests/integration/test_oom_canary/test.py new file mode 100644 index 000000000000..149f61b5f84d --- /dev/null +++ b/tests/integration/test_oom_canary/test.py @@ -0,0 +1,156 @@ +import threading +import time +import uuid +from dataclasses import dataclass + +impor...
[ "tests/integration/test_oom_canary/test.py" ]
[ { "comment": "The OOM response contract here includes more than `system.crash_log`: `onCanaryOOM` also calls `ProcessList::killAllQueries` and `MergeList::cancelAll`.\n\nCurrent coverage only proves relaunch + crash-log insertion, so a regression where query/merge cancellation stops working would still pass.\n\...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
`MADV_POPULATE_WRITE` is selected by compile-time macro, but support is kernel-runtime dependent. On older kernels (for example pre-5.14), `madvise(..., MADV_POPULATE_WRITE)` returns `EINVAL`; this path exits with `TRANSIENT`, and the monitor then burns through relaunch attempts and disables the canary permanently eve...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/programs/oom-canary/OomCanary.cpp b/programs/oom-canary/OomCanary.cpp new file mode 100644 index 000000000000..343410e486d6 --- /dev/null +++ b/programs/oom-canary/OomCanary.cpp @@ -0,0 +1,144 @@ +#include <Common/OOMCanaryExitCodes.h> + +#if defined(OS_LINUX) + +#include <Common/Exception.h> +#include <IO...
[ "programs/oom-canary/OomCanary.cpp" ]
[ { "comment": "`MADV_POPULATE_WRITE` is selected by compile-time macro, but support is kernel-runtime dependent.\n\nOn older kernels (for example pre-5.14), `madvise(..., MADV_POPULATE_WRITE)` returns `EINVAL`; this path exits with `TRANSIENT`, and the monitor then burns through relaunch attempts and disables th...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
`close_range` is gated only at compile time here. If the binary is built with `__NR_close_range` but runs on an older kernel where the syscall is unavailable, this call returns `ENOSYS` and we silently keep all inherited file descriptors open. That breaks the intended invariant that the canary process does not retain ...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/programs/oom-canary/OomCanary.cpp b/programs/oom-canary/OomCanary.cpp new file mode 100644 index 000000000000..343410e486d6 --- /dev/null +++ b/programs/oom-canary/OomCanary.cpp @@ -0,0 +1,144 @@ +#include <Common/OOMCanaryExitCodes.h> + +#if defined(OS_LINUX) + +#include <Common/Exception.h> +#include <IO...
[ "programs/oom-canary/OomCanary.cpp" ]
[ { "comment": "`close_range` is gated only at compile time here. If the binary is built with `__NR_close_range` but runs on an older kernel where the syscall is unavailable, this call returns `ENOSYS` and we silently keep all inherited file descriptors open.\n\nThat breaks the intended invariant that the canary ...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
`max_fd` is hard-capped to `65536`, so descriptors above that are never closed in the canary process. On deployments with high `max_open_files` (or simply many open descriptors), the canary can still inherit and keep parent sockets/files alive after relaunch, which breaks the isolation invariant and can retain resourc...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/programs/oom-canary/OomCanary.cpp b/programs/oom-canary/OomCanary.cpp new file mode 100644 index 000000000000..343410e486d6 --- /dev/null +++ b/programs/oom-canary/OomCanary.cpp @@ -0,0 +1,144 @@ +#include <Common/OOMCanaryExitCodes.h> + +#if defined(OS_LINUX) + +#include <Common/Exception.h> +#include <IO...
[ "programs/oom-canary/OomCanary.cpp" ]
[ { "comment": "`max_fd` is hard-capped to `65536`, so descriptors above that are never closed in the canary process.\n\nOn deployments with high `max_open_files` (or simply many open descriptors), the canary can still inherit and keep parent sockets/files alive after relaunch, which breaks the isolation invarian...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
Let's maybe log exception?
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/programs/oom-canary/OomCanary.cpp b/programs/oom-canary/OomCanary.cpp new file mode 100644 index 000000000000..343410e486d6 --- /dev/null +++ b/programs/oom-canary/OomCanary.cpp @@ -0,0 +1,144 @@ +#include <Common/OOMCanaryExitCodes.h> + +#if defined(OS_LINUX) + +#include <Common/Exception.h> +#include <IO...
[ "programs/oom-canary/OomCanary.cpp" ]
[ { "comment": "Let's maybe log exception?", "path": "programs/oom-canary/OomCanary.cpp", "hunk": "@@ -0,0 +1,112 @@\n+#include <Common/OOMCanaryExitCodes.h>\n+\n+#if defined(OS_LINUX)\n+\n+#include <IO/ReadHelpers.h>\n+#include <IO/WriteBufferFromFile.h>\n+#include <IO/WriteHelpers.h>\n+\n+#include <sys/...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
Let's log something
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/programs/oom-canary/OomCanary.cpp b/programs/oom-canary/OomCanary.cpp new file mode 100644 index 000000000000..343410e486d6 --- /dev/null +++ b/programs/oom-canary/OomCanary.cpp @@ -0,0 +1,144 @@ +#include <Common/OOMCanaryExitCodes.h> + +#if defined(OS_LINUX) + +#include <Common/Exception.h> +#include <IO...
[ "programs/oom-canary/OomCanary.cpp" ]
[ { "comment": "Let's log something", "path": "programs/oom-canary/OomCanary.cpp", "hunk": "@@ -0,0 +1,112 @@\n+#include <Common/OOMCanaryExitCodes.h>\n+\n+#if defined(OS_LINUX)\n+\n+#include <IO/ReadHelpers.h>\n+#include <IO/WriteBufferFromFile.h>\n+#include <IO/WriteHelpers.h>\n+\n+#include <sys/mman.h>...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
Let's log something
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/programs/oom-canary/OomCanary.cpp b/programs/oom-canary/OomCanary.cpp new file mode 100644 index 000000000000..343410e486d6 --- /dev/null +++ b/programs/oom-canary/OomCanary.cpp @@ -0,0 +1,144 @@ +#include <Common/OOMCanaryExitCodes.h> + +#if defined(OS_LINUX) + +#include <Common/Exception.h> +#include <IO...
[ "programs/oom-canary/OomCanary.cpp" ]
[ { "comment": "Let's log something", "path": "programs/oom-canary/OomCanary.cpp", "hunk": "@@ -0,0 +1,112 @@\n+#include <Common/OOMCanaryExitCodes.h>\n+\n+#if defined(OS_LINUX)\n+\n+#include <IO/ReadHelpers.h>\n+#include <IO/WriteBufferFromFile.h>\n+#include <IO/WriteHelpers.h>\n+\n+#include <sys/mman.h>...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
Do we need this lambda?
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/programs/server/Server.cpp b/programs/server/Server.cpp index 8343f9411dbc..c56d0593780f 100644 --- a/programs/server/Server.cpp +++ b/programs/server/Server.cpp @@ -31,6 +31,7 @@ #include <Common/CurrentMemoryTracker.h> #include <Common/MemoryTracker.h> #include <Common/MemoryWorker.h> +#include <Commo...
[ "programs/server/Server.cpp" ]
[ { "comment": "Do we need this lambda?", "path": "programs/server/Server.cpp", "hunk": "@@ -3252,6 +3296,8 @@ try\n else\n global_context->cancelAllBackupsAndRestores();\n \n+ stop_oom_canary();", "resolving_sha": "f6a6dce6c24bdd0f8a3630844943a0a43218c964", ...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
Why it should be in IO?
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt index ab469adee233..7e7318e148d3 100644 --- a/src/CMakeLists.txt +++ b/src/CMakeLists.txt @@ -153,6 +153,7 @@ add_headers_and_sources(dbms Storages/ObjectStorage/DataLakes/Iceberg) add_headers_and_sources(dbms Storages/ObjectStorage/DataLakes/DeltaLake) add_headers...
[ "src/CMakeLists.txt" ]
[ { "comment": "Why it should be in IO?", "path": "src/CMakeLists.txt", "hunk": "@@ -111,6 +111,8 @@ add_headers_and_sources(clickhouse_common_io IO/AzureBlobStorage)\n add_headers_and_sources(clickhouse_common_io Disks/DiskObjectStorage/ObjectStorages/AzureBlobStorage)\n list (REMOVE_ITEM clickhouse_comm...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
Why do we need both allow_experimental_oom_canary and oom_canary_enable? Looks a little bit overkill
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Core/ServerSettings.cpp b/src/Core/ServerSettings.cpp index ee0a3d9a0d13..a818c288cec6 100644 --- a/src/Core/ServerSettings.cpp +++ b/src/Core/ServerSettings.cpp @@ -1505,6 +1505,45 @@ The policy on how to perform a scheduling of CPU slots specified by `concurrent_ ``` )", 0) \ DECLARE(I...
[ "src/Core/ServerSettings.cpp" ]
[ { "comment": "Why do we need both allow_experimental_oom_canary and oom_canary_enable? Looks a little bit overkill", "path": "src/Core/ServerSettings.cpp", "hunk": "@@ -1495,6 +1495,38 @@ The policy on how to perform a scheduling of CPU slots specified by `concurrent_\n ```\n )\", 0) \\\n DE...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
And also make sense to log total memory usage
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Common/OOMCanary/OOMCanary.cpp b/src/Common/OOMCanary/OOMCanary.cpp new file mode 100644 index 000000000000..69f795a2ba0e --- /dev/null +++ b/src/Common/OOMCanary/OOMCanary.cpp @@ -0,0 +1,476 @@ +#include <Common/OOMCanary/OOMCanary.h> + +#if defined(OS_LINUX) + +#include <Common/Epoll.h> +#include <Co...
[ "src/Common/OOMCanary/OOMCanary.cpp" ]
[ { "comment": "And also make sense to log total memory usage", "path": "src/Common/OOMCanary/OOMCanary.cpp", "hunk": "@@ -0,0 +1,395 @@\n+#include <Common/OOMCanary.h>\n+\n+#if defined(OS_LINUX)\n+\n+#include <Common/Epoll.h>\n+#include <Common/OOMCanaryExitCodes.h>\n+#include <Common/logger_useful.h>\n+...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
Maybe it is enough to mark it ONCE, that wait you don't need to disable it
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Common/FailPoint.cpp b/src/Common/FailPoint.cpp index e5ff1302db09..d89c04afd094 100644 --- a/src/Common/FailPoint.cpp +++ b/src/Common/FailPoint.cpp @@ -179,6 +179,7 @@ static struct InitFiu REGULAR(check_database_datalake_negative) \ REGULAR(restart_replica_fail_after_detach) \ REGULAR...
[ "src/Common/FailPoint.cpp" ]
[ { "comment": "Maybe it is enough to mark it ONCE, that wait you don't need to disable it", "path": "src/Common/FailPoint.cpp", "hunk": "@@ -177,6 +177,7 @@ static struct InitFiu\n REGULAR(check_database_datalake_negative) \\\n REGULAR(restart_replica_fail_after_detach) \\\n REGULAR(database_...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
Also I guess it make sense to move this to the top
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/programs/oom-canary/OomCanary.cpp b/programs/oom-canary/OomCanary.cpp new file mode 100644 index 000000000000..343410e486d6 --- /dev/null +++ b/programs/oom-canary/OomCanary.cpp @@ -0,0 +1,144 @@ +#include <Common/OOMCanaryExitCodes.h> + +#if defined(OS_LINUX) + +#include <Common/Exception.h> +#include <IO...
[ "programs/oom-canary/OomCanary.cpp" ]
[ { "comment": "Also I guess it make sense to move this to the top", "path": "programs/oom-canary/OomCanary.cpp", "hunk": "@@ -0,0 +1,112 @@\n+#include <Common/OOMCanaryExitCodes.h>\n+\n+#if defined(OS_LINUX)\n+\n+#include <IO/ReadHelpers.h>\n+#include <IO/WriteBufferFromFile.h>\n+#include <IO/WriteHelper...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
The PR description/changelog text still references `allow_experimental_oom_canary`, but that setting is not present in the current code anymore (`src/Core/ServerSettings.cpp` only defines `oom_canary_enable` + related tuning settings). Please update the PR template fields so release notes/documentation match the shipp...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Core/ServerSettings.cpp b/src/Core/ServerSettings.cpp index ee0a3d9a0d13..a818c288cec6 100644 --- a/src/Core/ServerSettings.cpp +++ b/src/Core/ServerSettings.cpp @@ -1505,6 +1505,45 @@ The policy on how to perform a scheduling of CPU slots specified by `concurrent_ ``` )", 0) \ DECLARE(I...
[ "src/Core/ServerSettings.cpp" ]
[ { "comment": "The PR description/changelog text still references `allow_experimental_oom_canary`, but that setting is not present in the current code anymore (`src/Core/ServerSettings.cpp` only defines `oom_canary_enable` + related tuning settings).\n\nPlease update the PR template fields so release notes/docum...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
`oom_canary_max_rapid_relaunches` is documented as limiting consecutive relaunches under sustained memory pressure, but confirmed OOMs reset `attempt` in `OOMCanary::monitorThread` before the cap is checked. With `oom_canary_relaunch=1`, a canary killed by real cgroup OOM evidence repeatedly will always go through `att...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Core/ServerSettings.cpp b/src/Core/ServerSettings.cpp index ee0a3d9a0d13..a818c288cec6 100644 --- a/src/Core/ServerSettings.cpp +++ b/src/Core/ServerSettings.cpp @@ -1505,6 +1505,45 @@ The policy on how to perform a scheduling of CPU slots specified by `concurrent_ ``` )", 0) \ DECLARE(I...
[ "src/Core/ServerSettings.cpp" ]
[ { "comment": "`oom_canary_max_rapid_relaunches` is documented as limiting consecutive relaunches under sustained memory pressure, but confirmed OOMs reset `attempt` in `OOMCanary::monitorThread` before the cap is checked. With `oom_canary_relaunch=1`, a canary killed by real cgroup OOM evidence repeatedly will ...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
The OOM response is supposed to shed memory by cancelling running queries, but this call goes through the regular `ProcessList::killAllQueries` path, which first allocates/reserves a `std::vector<QueryStatusPtr>` proportional to the number of running queries before it cancels anything. Under the memory pressure that ju...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Common/OOMCanary/OOMCanary.cpp b/src/Common/OOMCanary/OOMCanary.cpp new file mode 100644 index 000000000000..69f795a2ba0e --- /dev/null +++ b/src/Common/OOMCanary/OOMCanary.cpp @@ -0,0 +1,476 @@ +#include <Common/OOMCanary/OOMCanary.h> + +#if defined(OS_LINUX) + +#include <Common/Epoll.h> +#include <Co...
[ "src/Common/OOMCanary/OOMCanary.cpp" ]
[ { "comment": "The OOM response is supposed to shed memory by cancelling running queries, but this call goes through the regular `ProcessList::killAllQueries` path, which first allocates/reserves a `std::vector<QueryStatusPtr>` proportional to the number of running queries before it cancels anything. Under the m...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
`Epoll::add` and the following `Epoll::getManyReady` can throw after the canary child has already been spawned. Because `monitorThread` has no top-level cleanup for that per-child state, the `ThreadFromGlobalPool` job exits, `stop` later only joins the finished job, and the child remains alive/unmonitored until parent ...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Common/OOMCanary/OOMCanary.cpp b/src/Common/OOMCanary/OOMCanary.cpp new file mode 100644 index 000000000000..69f795a2ba0e --- /dev/null +++ b/src/Common/OOMCanary/OOMCanary.cpp @@ -0,0 +1,476 @@ +#include <Common/OOMCanary/OOMCanary.h> + +#if defined(OS_LINUX) + +#include <Common/Epoll.h> +#include <Co...
[ "src/Common/OOMCanary/OOMCanary.cpp" ]
[ { "comment": "`Epoll::add` and the following `Epoll::getManyReady` can throw after the canary child has already been spawned. Because `monitorThread` has no top-level cleanup for that per-child state, the `ThreadFromGlobalPool` job exits, `stop` later only joins the finished job, and the child remains alive/unm...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
BTW I think we should try to use "getCurrentExceptionMessage" here, since std::exception will not have any details that the DB::Exception has
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/programs/oom-canary/OomCanary.cpp b/programs/oom-canary/OomCanary.cpp new file mode 100644 index 000000000000..343410e486d6 --- /dev/null +++ b/programs/oom-canary/OomCanary.cpp @@ -0,0 +1,144 @@ +#include <Common/OOMCanaryExitCodes.h> + +#if defined(OS_LINUX) + +#include <Common/Exception.h> +#include <IO...
[ "programs/oom-canary/OomCanary.cpp" ]
[ { "comment": "BTW I think we should try to use \"getCurrentExceptionMessage\" here, since std::exception will not have any details that the DB::Exception has", "path": "programs/oom-canary/OomCanary.cpp", "hunk": "@@ -0,0 +1,127 @@\n+#include <Common/OOMCanaryExitCodes.h>\n+\n+#if defined(OS_LINUX)\n+\n...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
```suggestion )", EXPERIMENTAL) \ ```
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Core/ServerSettings.cpp b/src/Core/ServerSettings.cpp index ee0a3d9a0d13..a818c288cec6 100644 --- a/src/Core/ServerSettings.cpp +++ b/src/Core/ServerSettings.cpp @@ -1505,6 +1505,45 @@ The policy on how to perform a scheduling of CPU slots specified by `concurrent_ ``` )", 0) \ DECLARE(I...
[ "src/Core/ServerSettings.cpp" ]
[ { "comment": "```suggestion\r\n )\", EXPERIMENTAL) \\\r\n```", "path": "src/Core/ServerSettings.cpp", "hunk": "@@ -1505,6 +1505,33 @@ The policy on how to perform a scheduling of CPU slots specified by `concurrent_\n ```\n )\", 0) \\\n DECLARE(Int32, oom_score, getDefaultOomScore(), R\"(O...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
`oom_canary_size` says the canary memory is allocated and locked, but the helper treats `mlock` as best-effort and ignores failures. With the default 100 MB size, many deployments will exceed `RLIMIT_MEMLOCK`, so operators can enable the canary believing the memory is locked when it is actually only touched and may be ...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Core/ServerSettings.cpp b/src/Core/ServerSettings.cpp index ee0a3d9a0d13..a818c288cec6 100644 --- a/src/Core/ServerSettings.cpp +++ b/src/Core/ServerSettings.cpp @@ -1505,6 +1505,45 @@ The policy on how to perform a scheduling of CPU slots specified by `concurrent_ ``` )", 0) \ DECLARE(I...
[ "src/Core/ServerSettings.cpp" ]
[ { "comment": "`oom_canary_size` says the canary memory is allocated and locked, but the helper treats `mlock` as best-effort and ignores failures. With the default 100 MB size, many deployments will exceed `RLIMIT_MEMLOCK`, so operators can enable the canary believing the memory is locked when it is actually on...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
`oom_canary_relaunch` documents only the confirmed OOM path, but `monitorThread` relaunches after any child death that does not hit a permanent setup failure: manual `SIGKILL` without cgroup evidence, non-`SIGKILL` signals, and `TRANSIENT` exits all fall through to the relaunch/backoff block. The integration test also ...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Core/ServerSettings.cpp b/src/Core/ServerSettings.cpp index ee0a3d9a0d13..a818c288cec6 100644 --- a/src/Core/ServerSettings.cpp +++ b/src/Core/ServerSettings.cpp @@ -1505,6 +1505,45 @@ The policy on how to perform a scheduling of CPU slots specified by `concurrent_ ``` )", 0) \ DECLARE(I...
[ "src/Core/ServerSettings.cpp" ]
[ { "comment": "`oom_canary_relaunch` documents only the confirmed OOM path, but `monitorThread` relaunches after any child death that does not hit a permanent setup failure: manual `SIGKILL` without cgroup evidence, non-`SIGKILL` signals, and `TRANSIENT` exits all fall through to the relaunch/backoff block. The ...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
The canary monitor should catch failures from the whole thread body, not only from `ThreadFromGlobalPool` scheduling. `monitorThread` starts by constructing `Epoll` and adding `shutdown_fd`; both operations can throw (`epoll_create1` / `epoll_ctl`, e.g. `EMFILE` or `ENOMEM`) before the per-child try/catch is reached. ...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Common/OOMCanary/OOMCanary.cpp b/src/Common/OOMCanary/OOMCanary.cpp new file mode 100644 index 000000000000..69f795a2ba0e --- /dev/null +++ b/src/Common/OOMCanary/OOMCanary.cpp @@ -0,0 +1,476 @@ +#include <Common/OOMCanary/OOMCanary.h> + +#if defined(OS_LINUX) + +#include <Common/Epoll.h> +#include <Co...
[ "src/Common/OOMCanary/OOMCanary.cpp" ]
[ { "comment": "The canary monitor should catch failures from the whole thread body, not only from `ThreadFromGlobalPool` scheduling.\n\n`monitorThread` starts by constructing `Epoll` and adding `shutdown_fd`; both operations can throw (`epoll_create1` / `epoll_ctl`, e.g. `EMFILE` or `ENOMEM`) before the per-chil...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
This makes the canary require `pidfd_open` support, but the user-facing contract only says Linux + cgroup v2 `memory.events`. On Linux kernels before `pidfd_open` (`ENOSYS`) or environments where it is blocked, enabling `oom_canary_enable` will spawn the helper, immediately kill it here, and disable the feature even t...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Common/OOMCanary/OOMCanary.cpp b/src/Common/OOMCanary/OOMCanary.cpp new file mode 100644 index 000000000000..69f795a2ba0e --- /dev/null +++ b/src/Common/OOMCanary/OOMCanary.cpp @@ -0,0 +1,476 @@ +#include <Common/OOMCanary/OOMCanary.h> + +#if defined(OS_LINUX) + +#include <Common/Epoll.h> +#include <Co...
[ "src/Common/OOMCanary/OOMCanary.cpp" ]
[ { "comment": "This makes the canary require `pidfd_open` support, but the user-facing contract only says Linux + cgroup v2 `memory.events`.\n\nOn Linux kernels before `pidfd_open` (`ENOSYS`) or environments where it is blocked, enabling `oom_canary_enable` will spawn the helper, immediately kill it here, and di...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
`oom_canary_max_rapid_relaunches` is documented as a rapid/consecutive relaunch guard, but this counter never resets after the canary has run stably. That makes it a lifetime death counter: for example, 11 unrelated OOMs or manual `SIGKILL`s spread over days will eventually disable the canary permanently, and the backo...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Common/OOMCanary/OOMCanary.cpp b/src/Common/OOMCanary/OOMCanary.cpp new file mode 100644 index 000000000000..69f795a2ba0e --- /dev/null +++ b/src/Common/OOMCanary/OOMCanary.cpp @@ -0,0 +1,476 @@ +#include <Common/OOMCanary/OOMCanary.h> + +#if defined(OS_LINUX) + +#include <Common/Epoll.h> +#include <Co...
[ "src/Common/OOMCanary/OOMCanary.cpp" ]
[ { "comment": "`oom_canary_max_rapid_relaunches` is documented as a rapid/consecutive relaunch guard, but this counter never resets after the canary has run stably. That makes it a lifetime death counter: for example, 11 unrelated OOMs or manual `SIGKILL`s spread over days will eventually disable the canary perm...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
The startup check accepts any cgroup v2 with `memory.events`, but the canary cannot protect the server when the same cgroup has `memory.oom.group=1`: on memcg OOM the kernel kills the whole cgroup as one unit, so the parent server is killed together with the sacrificial child and the monitor never reaches `onCanaryOOM`...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Common/OOMCanary/OOMCanary.cpp b/src/Common/OOMCanary/OOMCanary.cpp new file mode 100644 index 000000000000..69f795a2ba0e --- /dev/null +++ b/src/Common/OOMCanary/OOMCanary.cpp @@ -0,0 +1,476 @@ +#include <Common/OOMCanary/OOMCanary.h> + +#if defined(OS_LINUX) + +#include <Common/Epoll.h> +#include <Co...
[ "src/Common/OOMCanary/OOMCanary.cpp" ]
[ { "comment": "The startup check accepts any cgroup v2 with `memory.events`, but the canary cannot protect the server when the same cgroup has `memory.oom.group=1`: on memcg OOM the kernel kills the whole cgroup as one unit, so the parent server is killed together with the sacrificial child and the monitor never...
true
ClickHouse/ClickHouse
101,942
comment_to_fix
Add OOM canary: a sacrificial child process to protect the server from the OOM killer
`readOOMKillCounter` uses `memory.events` as the OOM discriminator, but in cgroup v2 `memory.events` is hierarchical by default: an OOM kill in a descendant cgroup also advances this `oom_kill` counter. That breaks the "cgroup-local evidence" contract here. A manual `SIGKILL` of the canary racing with an OOM in a child...
961aff5091dd55df14a173b7b3f58c1f9524b7e0
f6a6dce6c24bdd0f8a3630844943a0a43218c964
diff --git a/src/Common/OOMCanary/OOMCanary.cpp b/src/Common/OOMCanary/OOMCanary.cpp new file mode 100644 index 000000000000..69f795a2ba0e --- /dev/null +++ b/src/Common/OOMCanary/OOMCanary.cpp @@ -0,0 +1,476 @@ +#include <Common/OOMCanary/OOMCanary.h> + +#if defined(OS_LINUX) + +#include <Common/Epoll.h> +#include <Co...
[ "src/Common/OOMCanary/OOMCanary.cpp" ]
[ { "comment": "`readOOMKillCounter` uses `memory.events` as the OOM discriminator, but in cgroup v2 `memory.events` is hierarchical by default: an OOM kill in a descendant cgroup also advances this `oom_kill` counter. That breaks the \"cgroup-local evidence\" contract here. A manual `SIGKILL` of the canary racin...
true
ClickHouse/ClickHouse
107,420
issue_to_patch
Skip 02947_parallel_replicas_remote on macOS (arm_darwin)
<!-- Linked issues and pull requests. Use full GitHub URLs, one relationship per line; delete the lines you don't need. Related: https://github.com/ClickHouse/ClickHouse/pull/106689 --> Related: https://github.com/ClickHouse/ClickHouse/pull/106689 ### Changelog category (leave one): - CI Fix or Improvement (changelo...
bf2051399e17d9e2b620d7d2c02568ff125011a6
cd91d6e66e9de43f359a4b7fce79ad047ebf41a6
diff --git a/ci/defs/darwin.skip b/ci/defs/darwin.skip index d493e709843c..9bca9e93921e 100644 --- a/ci/defs/darwin.skip +++ b/ci/defs/darwin.skip @@ -1146,6 +1146,7 @@ 02947_merge_tree_index_table_1 02947_merge_tree_index_table_2 02947_merge_tree_index_table_4 +02947_parallel_replicas_remote 02950_dictionary_ssd_c...
[ "ci/defs/darwin.skip" ]
[]
true
ClickHouse/ClickHouse
106,178
issue_to_patch
Add embedded documentation for database engines
This is the second PR in the series that adds embedded, runtime-introspectable documentation to ClickHouse component registries. The first ([#106177](https://github.com/ClickHouse/ClickHouse/pull/106177)) did this for table engines; this one does it for database engines. What it does: - Attaches the shared `Documentat...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
64398d8d8d683085114dd085d2c89e005c18fea6
diff --git a/docs/en/engines/database-engines/backup.md b/docs/en/engines/database-engines/backup.md index 98fc17e402c3..cb3880d4c9c3 100644 --- a/docs/en/engines/database-engines/backup.md +++ b/docs/en/engines/database-engines/backup.md @@ -14,20 +14,13 @@ Database backup works with both incremental and non-increment...
[ "docs/en/engines/database-engines/backup.md", "docs/en/engines/database-engines/datalake.md", "docs/en/engines/database-engines/materialized-postgresql.md", "docs/en/operations/system-tables/database_engines.md", "src/Common/Documentation.h", "src/Databases/DataLake/DatabaseDataLake.cpp", "src/Databases...
[ { "comment": "The page now documents the new `description`, `syntax`, `examples`, `introduced_in`, and `related` columns, but the example below still runs `SELECT *` and shows only the old `name` column. That result is no longer possible, so readers copying the docs will see a different shape. Please either mak...
diff --git a/tests/queries/0_stateless/02959_system_database_engines.sql b/tests/queries/0_stateless/02959_system_database_engines.sql index 67cb20f04006..f973a1dbce51 100644 --- a/tests/queries/0_stateless/02959_system_database_engines.sql +++ b/tests/queries/0_stateless/02959_system_database_engines.sql @@ -1,1 +1,1 ...
true
ClickHouse/ClickHouse
106,178
comment_to_fix
Add embedded documentation for database engines
The page now documents the new `description`, `syntax`, `examples`, `introduced_in`, and `related` columns, but the example below still runs `SELECT *` and shows only the old `name` column. That result is no longer possible, so readers copying the docs will see a different shape. Please either make the example query se...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
64398d8d8d683085114dd085d2c89e005c18fea6
diff --git a/docs/en/operations/system-tables/database_engines.md b/docs/en/operations/system-tables/database_engines.md index 5f81c82f64c7..7ad26216b5ae 100644 --- a/docs/en/operations/system-tables/database_engines.md +++ b/docs/en/operations/system-tables/database_engines.md @@ -15,20 +15,25 @@ Contains the list of ...
[ "docs/en/operations/system-tables/database_engines.md" ]
[ { "comment": "The page now documents the new `description`, `syntax`, `examples`, `introduced_in`, and `related` columns, but the example below still runs `SELECT *` and shows only the old `name` column. That result is no longer possible, so readers copying the docs will see a different shape. Please either mak...
true
ClickHouse/ClickHouse
106,178
comment_to_fix
Add embedded documentation for database engines
The `related` column is documented as containing names of related database engines, but `MaterializedMySQL` is not registered in `DatabaseFactory` anywhere in this tree, so this value cannot be resolved back through `system.database_engines`. Please keep `related` limited to registered database engine names, or add a v...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
64398d8d8d683085114dd085d2c89e005c18fea6
diff --git a/src/Databases/MySQL/DatabaseMySQL.cpp b/src/Databases/MySQL/DatabaseMySQL.cpp index 0c798900f1b6..a9494fa2abd3 100644 --- a/src/Databases/MySQL/DatabaseMySQL.cpp +++ b/src/Databases/MySQL/DatabaseMySQL.cpp @@ -667,7 +667,176 @@ void registerDatabaseMySQL(DatabaseFactory & factory) .supports_settin...
[ "src/Databases/MySQL/DatabaseMySQL.cpp" ]
[ { "comment": "The `related` column is documented as containing names of related database engines, but `MaterializedMySQL` is not registered in `DatabaseFactory` anywhere in this tree, so this value cannot be resolved back through `system.database_engines`. Please keep `related` limited to registered database en...
true
ClickHouse/ClickHouse
106,178
comment_to_fix
Add embedded documentation for database engines
`Filesystem` is always registered, but both `S3` and `HDFS` are behind `USE_AWS_S3` / `USE_HDFS`. In builds without either option, `system.database_engines` will contain the `Filesystem` row with `related = ['S3','HDFS']`, while those related targets have no row in the same system table. The new test only validates alw...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
64398d8d8d683085114dd085d2c89e005c18fea6
diff --git a/src/Databases/DatabaseFilesystem.cpp b/src/Databases/DatabaseFilesystem.cpp index f85f41e81d93..0f6173cb8615 100644 --- a/src/Databases/DatabaseFilesystem.cpp +++ b/src/Databases/DatabaseFilesystem.cpp @@ -268,6 +268,9 @@ void registerDatabaseFilesystem(DatabaseFactory & factory) .supports_argumen...
[ "src/Databases/DatabaseFilesystem.cpp" ]
[ { "comment": "`Filesystem` is always registered, but both `S3` and `HDFS` are behind `USE_AWS_S3` / `USE_HDFS`. In builds without either option, `system.database_engines` will contain the `Filesystem` row with `related = ['S3','HDFS']`, while those related targets have no row in the same system table. The new t...
true
ClickHouse/ClickHouse
106,178
comment_to_fix
Add embedded documentation for database engines
This example still cannot produce the shown response. `Lazy` is not registered in this tree (the stateless test for this same filter expects only `Atomic` and `Ordinary`), and without `ORDER BY` the row order is not stable because `system.database_engines` is filled from an unordered factory map. Since this PR is alrea...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
64398d8d8d683085114dd085d2c89e005c18fea6
diff --git a/docs/en/operations/system-tables/database_engines.md b/docs/en/operations/system-tables/database_engines.md index 5f81c82f64c7..7ad26216b5ae 100644 --- a/docs/en/operations/system-tables/database_engines.md +++ b/docs/en/operations/system-tables/database_engines.md @@ -15,20 +15,25 @@ Contains the list of ...
[ "docs/en/operations/system-tables/database_engines.md" ]
[ { "comment": "This example still cannot produce the shown response. `Lazy` is not registered in this tree (the stateless test for this same filter expects only `Atomic` and `Ordinary`), and without `ORDER BY` the row order is not stable because `system.database_engines` is filled from an unordered factory map. ...
true
ClickHouse/ClickHouse
106,178
comment_to_fix
Add embedded documentation for database engines
The new `syntax` column should expose a structurally valid engine clause, but this value makes the second `Backup` argument a string literal. `parseArguments` passes that AST to `BackupInfo::fromAST`, which requires an `ASTFunction`, so `CREATE DATABASE ... ENGINE = Backup('db', 'backup_destination')` throws `Expected ...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
64398d8d8d683085114dd085d2c89e005c18fea6
diff --git a/src/Databases/DatabaseBackup.cpp b/src/Databases/DatabaseBackup.cpp index 74b63d8150db..a60fb82f166a 100644 --- a/src/Databases/DatabaseBackup.cpp +++ b/src/Databases/DatabaseBackup.cpp @@ -482,7 +482,106 @@ void registerDatabaseBackup(DatabaseFactory & factory) return std::make_shared<DatabaseBac...
[ "src/Databases/DatabaseBackup.cpp" ]
[ { "comment": "The new `syntax` column should expose a structurally valid engine clause, but this value makes the second `Backup` argument a string literal. `parseArguments` passes that AST to `BackupInfo::fromAST`, which requires an `ASTFunction`, so `CREATE DATABASE ... ENGINE = Backup('db', 'backup_destinatio...
true
ClickHouse/ClickHouse
106,178
comment_to_fix
Add embedded documentation for database engines
The `description` field is now user-visible through `system.database_engines`, but this first `Backup` example is still not a valid `CREATE DATABASE` statement: the second argument is a string literal, and `parseArguments` passes it to `BackupInfo::fromAST`, which requires an `ASTFunction` and throws `Expected function...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
64398d8d8d683085114dd085d2c89e005c18fea6
diff --git a/src/Databases/DatabaseBackup.cpp b/src/Databases/DatabaseBackup.cpp index 74b63d8150db..a60fb82f166a 100644 --- a/src/Databases/DatabaseBackup.cpp +++ b/src/Databases/DatabaseBackup.cpp @@ -482,7 +482,106 @@ void registerDatabaseBackup(DatabaseFactory & factory) return std::make_shared<DatabaseBac...
[ "src/Databases/DatabaseBackup.cpp" ]
[ { "comment": "The `description` field is now user-visible through `system.database_engines`, but this first `Backup` example is still not a valid `CREATE DATABASE` statement: the second argument is a string literal, and `parseArguments` passes it to `BackupInfo::fromAST`, which requires an `ASTFunction` and thr...
true
ClickHouse/ClickHouse
106,178
comment_to_fix
Add embedded documentation for database engines
The PR promises the dedicated database-engine pages are embedded into `system.database_engines`, but this `MySQL` copy is already stale: the current `docs/en/engines/database-engines/mysql.md` page includes `[SETTINGS enable_compression=0]` plus the `enable_compression` settings section, while the embedded `description...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
64398d8d8d683085114dd085d2c89e005c18fea6
diff --git a/src/Databases/MySQL/DatabaseMySQL.cpp b/src/Databases/MySQL/DatabaseMySQL.cpp index 0c798900f1b6..a9494fa2abd3 100644 --- a/src/Databases/MySQL/DatabaseMySQL.cpp +++ b/src/Databases/MySQL/DatabaseMySQL.cpp @@ -667,7 +667,176 @@ void registerDatabaseMySQL(DatabaseFactory & factory) .supports_settin...
[ "src/Databases/MySQL/DatabaseMySQL.cpp" ]
[ { "comment": "The PR promises the dedicated database-engine pages are embedded into `system.database_engines`, but this `MySQL` copy is already stale: the current `docs/en/engines/database-engines/mysql.md` page includes `[SETTINGS enable_compression=0]` plus the `enable_compression` settings section, while the...
true
ClickHouse/ClickHouse
106,178
comment_to_fix
Add embedded documentation for database engines
This example creates `postgres_db`, but the query immediately uses `postgresql_db.postgres_table`. Since this block is now exposed through `system.database_engines.description`, the copied example references a database that was never created. Please make the database name consistent here and in `docs/en/engines/databas...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
64398d8d8d683085114dd085d2c89e005c18fea6
diff --git a/src/Databases/PostgreSQL/DatabaseMaterializedPostgreSQL.cpp b/src/Databases/PostgreSQL/DatabaseMaterializedPostgreSQL.cpp index 4031b373efb3..b17234d69570 100644 --- a/src/Databases/PostgreSQL/DatabaseMaterializedPostgreSQL.cpp +++ b/src/Databases/PostgreSQL/DatabaseMaterializedPostgreSQL.cpp @@ -568,7 +56...
[ "src/Databases/PostgreSQL/DatabaseMaterializedPostgreSQL.cpp" ]
[ { "comment": "This example creates `postgres_db`, but the query immediately uses `postgresql_db.postgres_table`. Since this block is now exposed through `system.database_engines.description`, the copied example references a database that was never created. Please make the database name consistent here and in `d...
true
ClickHouse/ClickHouse
106,178
comment_to_fix
Add embedded documentation for database engines
The embedded `description` is now exposed through `system.database_engines`, but this `sql` block still cannot be copied as a valid query: the statement ends with `:` instead of `;`, and the response table is inside the same `sql` fence. Please terminate the query with `;` and move the response to a `text` fence here a...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
64398d8d8d683085114dd085d2c89e005c18fea6
diff --git a/src/Databases/DatabaseBackup.cpp b/src/Databases/DatabaseBackup.cpp index 74b63d8150db..a60fb82f166a 100644 --- a/src/Databases/DatabaseBackup.cpp +++ b/src/Databases/DatabaseBackup.cpp @@ -482,7 +482,106 @@ void registerDatabaseBackup(DatabaseFactory & factory) return std::make_shared<DatabaseBac...
[ "src/Databases/DatabaseBackup.cpp" ]
[ { "comment": "The embedded `description` is now exposed through `system.database_engines`, but this `sql` block still cannot be copied as a valid query: the statement ends with `:` instead of `;`, and the response table is inside the same `sql` fence. Please terminate the query with `;` and move the response to...
true
ClickHouse/ClickHouse
107,417
issue_to_patch
Fix undefined behavior on a non-finite timestamp in prometheusQuery
<!-- Linked issues and pull requests. Use full GitHub URLs, one relationship per line; delete the lines you don't need. --> No related open issue found (searched parseTimeSeriesTypes / prometheusQuery UBSan / getFromFloat). ### Changelog category (leave one): - Bug Fix (user-visible misbehavior in an official stable ...
6a756730987624515ad9c08daab2b7831d343ba1
f76710f11cd3d9f161f1bfda5c6dcab075e448ad
diff --git a/src/Parsers/Prometheus/parseTimeSeriesTypes.cpp b/src/Parsers/Prometheus/parseTimeSeriesTypes.cpp index 9f1f69e5241d..bc140b6cbadf 100644 --- a/src/Parsers/Prometheus/parseTimeSeriesTypes.cpp +++ b/src/Parsers/Prometheus/parseTimeSeriesTypes.cpp @@ -46,6 +46,14 @@ namespace template <is_decimal T> ...
[ "src/Parsers/Prometheus/parseTimeSeriesTypes.cpp", "tests/queries/0_stateless/04337_prometheus_query_nan_timestamp.reference", "tests/queries/0_stateless/04337_prometheus_query_nan_timestamp.sql" ]
[]
diff --git a/tests/queries/0_stateless/04337_prometheus_query_nan_timestamp.reference b/tests/queries/0_stateless/04337_prometheus_query_nan_timestamp.reference new file mode 100644 index 000000000000..0b06b5563f22 --- /dev/null +++ b/tests/queries/0_stateless/04337_prometheus_query_nan_timestamp.reference @@ -0,0 +1,1...
true
ClickHouse/ClickHouse
107,398
issue_to_patch
Backport #107208 to 26.5: Do not narrow UnionStep pipeline when with read in order
Original pull-request https://github.com/ClickHouse/ClickHouse/pull/107208 Cherry-pick pull-request https://github.com/ClickHouse/ClickHouse/pull/107397 This pull-request is a last step of an automated backporting. Treat it as a standard pull-request: look at the checks and resolve conflicts. Merge it only if you inte...
503b8321beb136b493f594dd79efaa00bcf6c6c4
cb0ec6adb2b8544935d15a7ee8842fc48273a28a
diff --git a/src/Core/Settings.cpp b/src/Core/Settings.cpp index 8c17611c7a15..4329fa1aa1a9 100644 --- a/src/Core/Settings.cpp +++ b/src/Core/Settings.cpp @@ -5927,10 +5927,10 @@ Set default mode in INTERSECT query. Possible values: empty string, 'ALL', 'DIST Set default mode in EXCEPT query. Possible values: empty st...
[ "src/Core/Settings.cpp", "src/Interpreters/InterpreterSelectWithUnionQuery.cpp", "src/Planner/Planner.cpp", "src/Processors/QueryPlan/Optimizations/applyOrder.cpp", "src/Processors/QueryPlan/Optimizations/filterPushDown.cpp", "src/Processors/QueryPlan/Optimizations/liftUpUnion.cpp", "src/Processors/Quer...
[]
diff --git a/tests/queries/0_stateless/04324_union_all_narrow_read_in_order.reference b/tests/queries/0_stateless/04324_union_all_narrow_read_in_order.reference new file mode 100644 index 000000000000..f7432d7d09a5 --- /dev/null +++ b/tests/queries/0_stateless/04324_union_all_narrow_read_in_order.reference @@ -0,0 +1,8...
true
ClickHouse/ClickHouse
104,558
issue_to_patch
order by + limit applied before arrayJoin ### Company or project name _No response_ ### Describe what's wrong It appears that "limit" is applied together with "order by" before "arrayJoin" is executed ### Does it reproduce on the most recent release? Yes ### How to reproduce Latest clickhouse version: I would ...
Do not lift `arrayJoin` above `SortingStep` (issue #82279)
The `tryExecuteFunctionsAfterSorting` optimization in `src/Processors/QueryPlan/Optimizations/liftUpFunctions.cpp` splits the post-sort expression into a "needed for sorting" part (computes the sort keys, stays below `SortingStep`) and an "unneeded for sorting" part (lifted above `SortingStep`). It is unsound to lift `...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
fab23475deb70684d0dbccc0d18a7e25678d8983
diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp index 70111d67b637..2de69143a2f0 100644 --- a/src/Interpreters/InterpreterSelectQuery.cpp +++ b/src/Interpreters/InterpreterSelectQuery.cpp @@ -369,6 +369,23 @@ InterpreterSelectQuery::~InterpreterSelectQuery() = defa...
[ "src/Interpreters/InterpreterSelectQuery.cpp", "src/Planner/PlannerJoinTree.cpp", "src/Processors/QueryPlan/Optimizations/liftUpFunctions.cpp", "src/Processors/QueryPlan/Optimizations/limitPushDown.cpp", "src/Processors/QueryPlan/Optimizations/optimizeLazyMaterialization.cpp", "src/Processors/QueryPlan/Op...
[ { "comment": "This still leaves a row-truncating sort path uncovered. `pushLimitByIntoSort` runs later and can attach a per-stream `LIMIT BY` hint to this same `SortingStep` while `getLimit()` stays zero, so this guard lets `arrayJoin` be lifted above a sort that will still drop rows before expansion. For examp...
diff --git a/tests/queries/0_stateless/01533_multiple_nested.sql b/tests/queries/0_stateless/01533_multiple_nested.sql index 12fd19228cda..b255002d385d 100644 --- a/tests/queries/0_stateless/01533_multiple_nested.sql +++ b/tests/queries/0_stateless/01533_multiple_nested.sql @@ -67,6 +67,9 @@ INSERT INTO nested SELECT n...
true
ClickHouse/ClickHouse
104,558
comment_to_fix
Do not lift `arrayJoin` above `SortingStep` (issue #82279)
This still leaves a row-truncating sort path uncovered. `pushLimitByIntoSort` runs later and can attach a per-stream `LIMIT BY` hint to this same `SortingStep` while `getLimit()` stays zero, so this guard lets `arrayJoin` be lifted above a sort that will still drop rows before expansion. For example, with source rows o...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
fab23475deb70684d0dbccc0d18a7e25678d8983
diff --git a/src/Processors/QueryPlan/Optimizations/liftUpFunctions.cpp b/src/Processors/QueryPlan/Optimizations/liftUpFunctions.cpp index 68caafdb1b16..c43296be1aff 100644 --- a/src/Processors/QueryPlan/Optimizations/liftUpFunctions.cpp +++ b/src/Processors/QueryPlan/Optimizations/liftUpFunctions.cpp @@ -87,6 +87,22 @...
[ "src/Processors/QueryPlan/Optimizations/liftUpFunctions.cpp" ]
[ { "comment": "This still leaves a row-truncating sort path uncovered. `pushLimitByIntoSort` runs later and can attach a per-stream `LIMIT BY` hint to this same `SortingStep` while `getLimit()` stays zero, so this guard lets `arrayJoin` be lifted above a sort that will still drop rows before expansion. For examp...
true
ClickHouse/ClickHouse
104,558
comment_to_fix
Do not lift `arrayJoin` above `SortingStep` (issue #82279)
This regression only proves the new guard when the `MergeTree` read still has multiple streams: the bad plan needs the per-stream `LimitByTransform` to run before the lifted `arrayJoin`. The test creates four parts, but background merges are still enabled, so an unfixed build can become a false negative if those parts ...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
fab23475deb70684d0dbccc0d18a7e25678d8983
diff --git a/tests/queries/0_stateless/04318_arrayjoin_after_sort_with_limit_by.sql b/tests/queries/0_stateless/04318_arrayjoin_after_sort_with_limit_by.sql new file mode 100644 index 000000000000..607f77415449 --- /dev/null +++ b/tests/queries/0_stateless/04318_arrayjoin_after_sort_with_limit_by.sql @@ -0,0 +1,42 @@ +...
[ "tests/queries/0_stateless/04318_arrayjoin_after_sort_with_limit_by.sql" ]
[ { "comment": "This regression only proves the new guard when the `MergeTree` read still has multiple streams: the bad plan needs the per-stream `LimitByTransform` to run before the lifted `arrayJoin`. The test creates four parts, but background merges are still enabled, so an unfixed build can become a false ne...
true
ClickHouse/ClickHouse
104,558
comment_to_fix
Do not lift `arrayJoin` above `SortingStep` (issue #82279)
This no-join regression still goes through `SortingStep`, so it does not cover the same row-truncation invariant for a plain `LIMIT` above `arrayJoin`. The second-pass `optimizePrimaryKeyConditionAndLimit` walks from a `SourceStepWithFilterBase` through `ExpressionStep`s and calls `setLimit` when it sees the outer `Lim...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
fab23475deb70684d0dbccc0d18a7e25678d8983
diff --git a/tests/queries/0_stateless/04220_arrayjoin_after_sort_with_limit.sql b/tests/queries/0_stateless/04220_arrayjoin_after_sort_with_limit.sql new file mode 100644 index 000000000000..3e1e2f1a870d --- /dev/null +++ b/tests/queries/0_stateless/04220_arrayjoin_after_sort_with_limit.sql @@ -0,0 +1,63 @@ +-- Regres...
[ "tests/queries/0_stateless/04220_arrayjoin_after_sort_with_limit.sql" ]
[ { "comment": "This no-join regression still goes through `SortingStep`, so it does not cover the same row-truncation invariant for a plain `LIMIT` above `arrayJoin`. The second-pass `optimizePrimaryKeyConditionAndLimit` walks from a `SourceStepWithFilterBase` through `ExpressionStep`s and calls `setLimit` when ...
true
ClickHouse/ClickHouse
104,558
comment_to_fix
Do not lift `arrayJoin` above `SortingStep` (issue #82279)
This only checks the select list, but `ARRAY JOIN` clauses are stored separately in `ASTSelectQuery::arrayJoinExpressionList`, so the direct source-limit path can still cap the source before expansion. For example, `SELECT number FROM numbers(100) ARRAY JOIN if(number < 3, [], [number]) AS x LIMIT 3` should skip the em...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
fab23475deb70684d0dbccc0d18a7e25678d8983
diff --git a/src/Processors/QueryPlan/numbersLikeUtils.cpp b/src/Processors/QueryPlan/numbersLikeUtils.cpp index 8536d181b9d7..55e6221a3bd3 100644 --- a/src/Processors/QueryPlan/numbersLikeUtils.cpp +++ b/src/Processors/QueryPlan/numbersLikeUtils.cpp @@ -4,6 +4,7 @@ #include <Core/Settings.h> #include <Interpreters...
[ "src/Processors/QueryPlan/numbersLikeUtils.cpp" ]
[ { "comment": "This only checks the select list, but `ARRAY JOIN` clauses are stored separately in `ASTSelectQuery::arrayJoinExpressionList`, so the direct source-limit path can still cap the source before expansion. For example, `SELECT number FROM numbers(100) ARRAY JOIN if(number < 3, [], [number]) AS x LIMIT...
true
ClickHouse/ClickHouse
104,558
comment_to_fix
Do not lift `arrayJoin` above `SortingStep` (issue #82279)
This guard still misses `arrayJoin` when it is hidden behind a top-level `WITH` alias, because it only walks `query.select()` plus the explicit `ARRAY JOIN` clause. `ApplyWithAliasVisitor` keeps the alias expression in `query.with()`, so `WITH arrayJoin(if(number < 3, [], [number])) AS x SELECT x FROM numbers(100) LIMI...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
fab23475deb70684d0dbccc0d18a7e25678d8983
diff --git a/src/Processors/QueryPlan/numbersLikeUtils.cpp b/src/Processors/QueryPlan/numbersLikeUtils.cpp index 8536d181b9d7..55e6221a3bd3 100644 --- a/src/Processors/QueryPlan/numbersLikeUtils.cpp +++ b/src/Processors/QueryPlan/numbersLikeUtils.cpp @@ -4,6 +4,7 @@ #include <Core/Settings.h> #include <Interpreters...
[ "src/Processors/QueryPlan/numbersLikeUtils.cpp" ]
[ { "comment": "This guard still misses `arrayJoin` when it is hidden behind a top-level `WITH` alias, because it only walks `query.select()` plus the explicit `ARRAY JOIN` clause. `ApplyWithAliasVisitor` keeps the alias expression in `query.with()`, so `WITH arrayJoin(if(number < 3, [], [number])) AS x SELECT x ...
true
ClickHouse/ClickHouse
104,558
comment_to_fix
Do not lift `arrayJoin` above `SortingStep` (issue #82279)
This still leaves another hard source-limit path uncovered. The new guard stops the `numbers` / `primes` construction-time limit, but simple queries also set `query_info.trivial_limit` through `InterpreterSelectQuery::maxBlockSizeByLimit` without checking whether the current `SELECT` contains `arrayJoin`. `StorageLoop`...
3659e5601eac2078ed0f3ccef17d3e9ea9bebc3c
fab23475deb70684d0dbccc0d18a7e25678d8983
diff --git a/src/Processors/QueryPlan/numbersLikeUtils.cpp b/src/Processors/QueryPlan/numbersLikeUtils.cpp index 8536d181b9d7..55e6221a3bd3 100644 --- a/src/Processors/QueryPlan/numbersLikeUtils.cpp +++ b/src/Processors/QueryPlan/numbersLikeUtils.cpp @@ -4,6 +4,7 @@ #include <Core/Settings.h> #include <Interpreters...
[ "src/Processors/QueryPlan/numbersLikeUtils.cpp" ]
[ { "comment": "This still leaves another hard source-limit path uncovered. The new guard stops the `numbers` / `primes` construction-time limit, but simple queries also set `query_info.trivial_limit` through `InterpreterSelectQuery::maxBlockSizeByLimit` without checking whether the current `SELECT` contains `arr...
true
ClickHouse/ClickHouse
106,621
issue_to_patch
Fix `Memory` table abort on `ALTER` commands with no per-row data effect (STID 2380-3043)
Fixes `STID: 2380-3043` (`Logical error: Mutation of \`Memory\` table produced incomplete output: block 0 has 0 rows, expected N`), reported by @alexey-milovidov on https://github.com/ClickHouse/ClickHouse/pull/106615 as a chronic `BuzzHouse` finding hitting master and 8+ unrelated PRs. Per @alexey-milovidov directive...
e78a4dac750cb4e71fb1928c6e3d760817d45e78
a2562ae47772b1636ed6b807d67418b4153ea9a0
diff --git a/src/Storages/StorageMemory.cpp b/src/Storages/StorageMemory.cpp index f0fc30d478b6..065a3f4fa691 100644 --- a/src/Storages/StorageMemory.cpp +++ b/src/Storages/StorageMemory.cpp @@ -249,24 +249,38 @@ void StorageMemory::mutate(const MutationCommands & commands, ContextPtr context { std::lock_guard lo...
[ "src/Storages/StorageMemory.cpp", "tests/queries/0_stateless/04322_memory_mutate_no_op_commands_2380_3043.reference", "tests/queries/0_stateless/04322_memory_mutate_no_op_commands_2380_3043.sql" ]
[ { "comment": "This `MATERIALIZE_TTL` entry does not actually make `ALTER TABLE ... MATERIALIZE TTL` a no-op for `Memory`: `InterpreterAlterQuery::runCommandSegments` builds a `MutationsInterpreter` for validation before `StorageMemory::mutate` is called, and `MutationsInterpreter::prepare` still throws `Cannot ...
diff --git a/tests/queries/0_stateless/04322_memory_mutate_no_op_commands_2380_3043.reference b/tests/queries/0_stateless/04322_memory_mutate_no_op_commands_2380_3043.reference new file mode 100644 index 000000000000..d0b25890f90e --- /dev/null +++ b/tests/queries/0_stateless/04322_memory_mutate_no_op_commands_2380_304...
true
ClickHouse/ClickHouse
106,621
comment_to_fix
Fix `Memory` table abort on `ALTER` commands with no per-row data effect (STID 2380-3043)
This `MATERIALIZE_TTL` entry does not actually make `ALTER TABLE ... MATERIALIZE TTL` a no-op for `Memory`: `InterpreterAlterQuery::runCommandSegments` builds a `MutationsInterpreter` for validation before `StorageMemory::mutate` is called, and `MutationsInterpreter::prepare` still throws `Cannot MATERIALIZE TTL as the...
e78a4dac750cb4e71fb1928c6e3d760817d45e78
a2562ae47772b1636ed6b807d67418b4153ea9a0
diff --git a/src/Storages/StorageMemory.cpp b/src/Storages/StorageMemory.cpp index f0fc30d478b6..065a3f4fa691 100644 --- a/src/Storages/StorageMemory.cpp +++ b/src/Storages/StorageMemory.cpp @@ -249,24 +249,38 @@ void StorageMemory::mutate(const MutationCommands & commands, ContextPtr context { std::lock_guard lo...
[ "src/Storages/StorageMemory.cpp" ]
[ { "comment": "This `MATERIALIZE_TTL` entry does not actually make `ALTER TABLE ... MATERIALIZE TTL` a no-op for `Memory`: `InterpreterAlterQuery::runCommandSegments` builds a `MutationsInterpreter` for validation before `StorageMemory::mutate` is called, and `MutationsInterpreter::prepare` still throws `Cannot ...
true
ClickHouse/ClickHouse
106,621
comment_to_fix
Fix `Memory` table abort on `ALTER` commands with no per-row data effect (STID 2380-3043)
These `DROP_*` entries are not reached by user `ALTER` queries on `Memory`. `DROP INDEX`, `DROP PROJECTION`, and `DROP STATISTICS` are parsed into an `AlterCommands` segment, and `StorageMemory::checkAlterIsPossible` rejects those command types before `StorageMemory::mutate` is called. So the PR/changelog promise that ...
e78a4dac750cb4e71fb1928c6e3d760817d45e78
a2562ae47772b1636ed6b807d67418b4153ea9a0
diff --git a/src/Storages/StorageMemory.cpp b/src/Storages/StorageMemory.cpp index f0fc30d478b6..065a3f4fa691 100644 --- a/src/Storages/StorageMemory.cpp +++ b/src/Storages/StorageMemory.cpp @@ -249,24 +249,38 @@ void StorageMemory::mutate(const MutationCommands & commands, ContextPtr context { std::lock_guard lo...
[ "src/Storages/StorageMemory.cpp" ]
[ { "comment": "These `DROP_*` entries are not reached by user `ALTER` queries on `Memory`. `DROP INDEX`, `DROP PROJECTION`, and `DROP STATISTICS` are parsed into an `AlterCommands` segment, and `StorageMemory::checkAlterIsPossible` rejects those command types before `StorageMemory::mutate` is called. So the PR/c...
true
ClickHouse/ClickHouse
107,419
issue_to_patch
Retry the changed-files GitHub API call and report failures clearly
<!-- Related: https://github.com/ClickHouse/ClickHouse/pull/74691 --> `GH.get_changed_files` fetches the list of files a PR touches via `gh api .../pulls/<n>/files`. The `store_data` pre-hook fetches it once and stores it into the job KV data; several other workflow hooks (`pr_labels_and_category`, `team_notifications...
1a5e24c889955eaff7e2e29e19e0074f30bf1322
653085058b9b2b5f932c03146f3ed39565d3225f
diff --git a/ci/jobs/scripts/workflow_hooks/feature_docs.py b/ci/jobs/scripts/workflow_hooks/feature_docs.py index 11e1bd6488b7..f03428fbdb42 100644 --- a/ci/jobs/scripts/workflow_hooks/feature_docs.py +++ b/ci/jobs/scripts/workflow_hooks/feature_docs.py @@ -42,6 +42,11 @@ def check_docs(): info = Info() if L...
[ "ci/jobs/scripts/workflow_hooks/feature_docs.py", "ci/jobs/scripts/workflow_hooks/pr_labels_and_category.py", "ci/jobs/scripts/workflow_hooks/team_notifications.py", "ci/praktika/gh.py" ]
[ { "comment": "This is the contract the PR fixes, but it is not covered by a focused regression test. A future change could reintroduce the original one-shot behavior and CI would only catch it during the next transient GitHub outage. Please add a small `ci/tests/test_gh.py` test that monkeypatches `Info` and `S...
true
ClickHouse/ClickHouse
107,419
comment_to_fix
Retry the changed-files GitHub API call and report failures clearly
This is the contract the PR fixes, but it is not covered by a focused regression test. A future change could reintroduce the original one-shot behavior and CI would only catch it during the next transient GitHub outage. Please add a small `ci/tests/test_gh.py` test that monkeypatches `Info` and `Shell.get_res_stdout_st...
1a5e24c889955eaff7e2e29e19e0074f30bf1322
653085058b9b2b5f932c03146f3ed39565d3225f
diff --git a/ci/praktika/gh.py b/ci/praktika/gh.py index 90b67d88ced4..6646239ebccc 100644 --- a/ci/praktika/gh.py +++ b/ci/praktika/gh.py @@ -91,18 +91,30 @@ def get_changed_files(cls, strict=False) -> List[str]: # HEAD, same as deleted files, which were always included. jq_both_sides = ".filename, (...
[ "ci/praktika/gh.py" ]
[ { "comment": "This is the contract the PR fixes, but it is not covered by a focused regression test. A future change could reintroduce the original one-shot behavior and CI would only catch it during the next transient GitHub outage. Please add a small `ci/tests/test_gh.py` test that monkeypatches `Info` and `S...
true
ClickHouse/ClickHouse
107,229
issue_to_patch
Pass-through row count estimation for row-preserving transforming steps
Add pass-through for generic transform which preserve row count From #106955 ### Changelog category (leave one): - Performance Improvement ### Changelog entry (a [user-readable short description](https://github.com/ClickHouse/ClickHouse/blob/master/docs/changelog_entry_guidelines.md) of the changes that goes ...
09c2cbb60368d15a866900aea3c073fa633825a2
890ce94bdcfb0cf2cfebe197e326e9f9d1bcb7c9
diff --git a/src/Processors/QueryPlan/Optimizations/optimizeJoin.cpp b/src/Processors/QueryPlan/Optimizations/optimizeJoin.cpp index c44104df03f4..d37f4bf3a162 100644 --- a/src/Processors/QueryPlan/Optimizations/optimizeJoin.cpp +++ b/src/Processors/QueryPlan/Optimizations/optimizeJoin.cpp @@ -420,11 +420,16 @@ static ...
[ "src/Processors/QueryPlan/Optimizations/optimizeJoin.cpp" ]
[ { "comment": "Can't it change number of distinct values?", "path": "src/Processors/QueryPlan/Optimizations/optimizeJoin.cpp", "hunk": "@@ -420,11 +420,16 @@ static RelationStats estimateReadRowsCount(QueryPlan::Node & node, const Actions\n }\n return stats;\n }\n+\n #if CLICKHOUSE_CL...
true
ClickHouse/ClickHouse
107,388
issue_to_patch
Data race on ServerSettings during reload ### Describe the bug (From Claude) This is a unsynchronized read/write race on the global ServerSettings object that lives inside ContextSharedPart. Here's what's happening: The writer (thread T812, ConfigReloader): During a config reload, Server.cpp:2482 calls Context::s...
Revert "Revert "Make getServerSetting return live values for runtime-changeable settings""
Re-applies #105172 ("Make `getServerSetting` return live values for runtime-changeable settings"), which was reverted by #106930 because of a data race. `getServerSetting` read the live value of runtime-changeable server settings by copy-constructing the whole `ServerSettings` struct from the reference returned by `Co...
19059120f5e793db617f4370f6493365992fe4df
02ce0c53642fe63bae3648444a0ee19fea98fffb
diff --git a/src/Core/ServerSettings.cpp b/src/Core/ServerSettings.cpp index 664af43c27c8..4c5319cd8ead 100644 --- a/src/Core/ServerSettings.cpp +++ b/src/Core/ServerSettings.cpp @@ -30,9 +30,13 @@ #include <Common/MemoryTracker.h> #include <Common/DNSResolver.h> +#include <Common/ZooKeeper/ZooKeeper.h> +#include <...
[ "src/Core/ServerSettings.cpp", "src/Core/ServerSettings.h", "src/Functions/getServerSetting.cpp", "src/Interpreters/Context.cpp", "src/Interpreters/Context.h", "src/Storages/System/StorageSystemServerSettings.cpp", "tests/queries/0_stateless/04241_get_server_setting_live_values.reference", "tests/quer...
[ { "comment": "This restores the unsynchronized `ServerSettings` snapshot that triggered `#106902`. `Context::getServerSettings` returns a reference to `shared->server_settings` without taking `shared->mutex`, but config reload still mutates that same object under the mutex via `Context::setS3QueueDisableStreami...
diff --git a/tests/queries/0_stateless/04241_get_server_setting_live_values.reference b/tests/queries/0_stateless/04241_get_server_setting_live_values.reference new file mode 100644 index 000000000000..627e1097cda3 --- /dev/null +++ b/tests/queries/0_stateless/04241_get_server_setting_live_values.reference @@ -0,0 +1,5...
true
ClickHouse/ClickHouse
107,388
comment_to_fix
Revert "Revert "Make getServerSetting return live values for runtime-changeable settings""
This restores the unsynchronized `ServerSettings` snapshot that triggered `#106902`. `Context::getServerSettings` returns a reference to `shared->server_settings` without taking `shared->mutex`, but config reload still mutates that same object under the mutex via `Context::setS3QueueDisableStreaming` and `Context::setM...
19059120f5e793db617f4370f6493365992fe4df
02ce0c53642fe63bae3648444a0ee19fea98fffb
diff --git a/src/Functions/getServerSetting.cpp b/src/Functions/getServerSetting.cpp index 8cdf1110564a..290f54c76d6e 100644 --- a/src/Functions/getServerSetting.cpp +++ b/src/Functions/getServerSetting.cpp @@ -70,7 +70,21 @@ class FunctionGetServerSetting final : public IFunction std::string_view setting_na...
[ "src/Functions/getServerSetting.cpp" ]
[ { "comment": "This restores the unsynchronized `ServerSettings` snapshot that triggered `#106902`. `Context::getServerSettings` returns a reference to `shared->server_settings` without taking `shared->mutex`, but config reload still mutates that same object under the mutex via `Context::setS3QueueDisableStreami...
true
ClickHouse/ClickHouse
93,830
issue_to_patch
Support for ALTER TABLE ADD ENUM VALUES
### Changelog category (leave one): - New Feature ### Changelog entry (a [user-readable short description](https://github.com/ClickHouse/ClickHouse/blob/master/docs/changelog_entry_guidelines.md) of the changes that goes into CHANGELOG.md): Implement `ADD ENUM VALUES` in `ALTER TABLE` queries to simplify appending...
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/docs/en/sql-reference/data-types/enum.md b/docs/en/sql-reference/data-types/enum.md index 5f330bcd8766..ecbc90e8bc42 100644 --- a/docs/en/sql-reference/data-types/enum.md +++ b/docs/en/sql-reference/data-types/enum.md @@ -162,3 +162,26 @@ Enum values are also convertible to numeric types using the `toT` fu...
[ "docs/en/sql-reference/data-types/enum.md", "docs/en/sql-reference/statements/alter/column.md", "src/DataTypes/DataTypeEnum.cpp", "src/DataTypes/DataTypeEnum.h", "src/DataTypes/EnumValues.cpp", "src/DataTypes/EnumValues.h", "src/Parsers/ASTAlterQuery.cpp", "src/Parsers/ASTAlterQuery.h", "src/Parsers...
[ { "comment": "I think about syntax. We never had function in `expression` of MODIFY COLUMN\r\nhttps://clickhouse.com/docs/sql-reference/statements/alter/column#modify-column. \r\n\r\nHave you seen any examples of similar logic in other DBMS?", "path": "docs/en/sql-reference/data-types/enum.md", "hunk": ...
diff --git a/tests/queries/0_stateless/04307_add_to_enum.reference b/tests/queries/0_stateless/04307_add_to_enum.reference new file mode 100644 index 000000000000..39a4e9ed5ed6 --- /dev/null +++ b/tests/queries/0_stateless/04307_add_to_enum.reference @@ -0,0 +1,92 @@ +======= NULLABLE =============== +======= Enum ==...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
```suggestion There is a syntax sugar to add new values to enum using ALTER [MODIFY COLUMN ADD ENUM VALUES](../../sql-reference/statements/alter/column.md#modify-column-add-enum-values) ```
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/docs/en/sql-reference/data-types/enum.md b/docs/en/sql-reference/data-types/enum.md index 5f330bcd8766..ecbc90e8bc42 100644 --- a/docs/en/sql-reference/data-types/enum.md +++ b/docs/en/sql-reference/data-types/enum.md @@ -162,3 +162,26 @@ Enum values are also convertible to numeric types using the `toT` fu...
[ "docs/en/sql-reference/data-types/enum.md" ]
[ { "comment": "```suggestion\nThere is a syntax sugar to add new values to enum using ALTER [MODIFY COLUMN ADD ENUM VALUES](../../sql-reference/statements/alter/column.md#modify-column-add-enum-values)\n```", "path": "docs/en/sql-reference/data-types/enum.md", "hunk": "@@ -164,3 +164,26 @@ Enum values ar...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
Typo: `is_nullale` β†’ `is_nullable`
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/Storages/AlterCommands.cpp b/src/Storages/AlterCommands.cpp index 3863adc67515..387e0f5e55c2 100644 --- a/src/Storages/AlterCommands.cpp +++ b/src/Storages/AlterCommands.cpp @@ -81,6 +81,7 @@ namespace ErrorCodes extern const int NOT_IMPLEMENTED; extern const int ALTER_OF_COLUMN_IS_FORBIDDEN;...
[ "src/Storages/AlterCommands.cpp" ]
[ { "comment": "Typo: `is_nullale` β†’ `is_nullable`", "path": "src/Storages/AlterCommands.cpp", "hunk": "@@ -1363,6 +1373,76 @@ void AlterCommands::prepare(const StorageInMemoryMetadata & metadata)\n command.default_expression = column_from_table.default_desc.expression;\n ...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
Perhaps it would be great to also add an existing enum value (e.g. `'Four' = 4`) to verify it's a noop.
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/tests/queries/0_stateless/04307_add_to_enum.sql.j2 b/tests/queries/0_stateless/04307_add_to_enum.sql.j2 new file mode 100644 index 000000000000..250dd7fdf109 --- /dev/null +++ b/tests/queries/0_stateless/04307_add_to_enum.sql.j2 @@ -0,0 +1,129 @@ +{% for NULLABLE_START, NULLABLE_END in {'' : '', 'Nullable(...
[ "tests/queries/0_stateless/04307_add_to_enum.sql.j2" ]
[ { "comment": "Perhaps it would be great to also add an existing enum value (e.g. `'Four' = 4`) to verify it's a noop.", "path": "tests/queries/0_stateless/04307_add_to_enum.sql.j2", "hunk": "@@ -0,0 +1,45 @@\n+{% for NULLABLE_START, NULLABLE_END in {'' : '', 'Nullable(' : ')'}.items() -%}\n+SELECT '====...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
`add_enum_values` is cloned and formatted, but `ASTAlterCommand::forEachPointerToChild` is not updated for it. During generic AST rewrites, `children` may be replaced while this typed field stays stale, which can desynchronize the command state. Please add `f(nullptr, &add_enum_values)` (and, for consistency, `f(nullpt...
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/Parsers/ASTAlterQuery.cpp b/src/Parsers/ASTAlterQuery.cpp index e0a808c98056..17a8e6ca3420 100644 --- a/src/Parsers/ASTAlterQuery.cpp +++ b/src/Parsers/ASTAlterQuery.cpp @@ -67,6 +67,8 @@ ASTPtr ASTAlterCommand::clone() const res->rename_to = res->children.emplace_back(rename_to->clone()).get(...
[ "src/Parsers/ASTAlterQuery.cpp" ]
[ { "comment": "`add_enum_values` is cloned and formatted, but `ASTAlterCommand::forEachPointerToChild` is not updated for it. During generic AST rewrites, `children` may be replaced while this typed field stays stale, which can desynchronize the command state. Please add `f(nullptr, &add_enum_values)` (and, for ...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
### ADD ENUM VALUES silently drops column default expressions **High Severity** <!-- DESCRIPTION START --> The `add_enum_values` processing sets `command.data_type` *after* the default-expression preservation check. The preservation check at the start of the `has_column` block (`if (command.data_type && !command.defa...
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/Storages/AlterCommands.cpp b/src/Storages/AlterCommands.cpp index 3863adc67515..387e0f5e55c2 100644 --- a/src/Storages/AlterCommands.cpp +++ b/src/Storages/AlterCommands.cpp @@ -81,6 +81,7 @@ namespace ErrorCodes extern const int NOT_IMPLEMENTED; extern const int ALTER_OF_COLUMN_IS_FORBIDDEN;...
[ "src/Storages/AlterCommands.cpp" ]
[ { "comment": "### ADD ENUM VALUES silently drops column default expressions\n\n**High Severity**\n\n<!-- DESCRIPTION START -->\nThe `add_enum_values` processing sets `command.data_type` *after* the default-expression preservation check. The preservation check at the start of the `has_column` block (`if (command...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
### Overflow error message reports wrong value for relative enums **Medium Severity** <!-- DESCRIPTION START --> In `mergeEnumTypes`, when `is_relative` is true, the overflow error message reports `val` (the original auto-assigned value, e.g., 1) instead of `candidate64` (the actual computed value, e.g., 128). A user...
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/DataTypes/DataTypeEnum.cpp b/src/DataTypes/DataTypeEnum.cpp index 682cb3106b26..e401538cfdc8 100644 --- a/src/DataTypes/DataTypeEnum.cpp +++ b/src/DataTypes/DataTypeEnum.cpp @@ -16,6 +16,7 @@ #include <Interpreters/Context.h> #include <Core/Settings.h> +#include <algorithm> #include <limits> @...
[ "src/DataTypes/DataTypeEnum.cpp" ]
[ { "comment": "### Overflow error message reports wrong value for relative enums\n\n**Medium Severity**\n\n<!-- DESCRIPTION START -->\nIn `mergeEnumTypes`, when `is_relative` is true, the overflow error message reports `val` (the original auto-assigned value, e.g., 1) instead of `candidate64` (the actual compute...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
When `ADD ENUM VALUES` sets `command.data_type` here, `default_kind`/`default_expression` are still unset for this command. Later, `AlterCommand::apply` treats any `data_type` change as a default rewrite (`if (default_expression || data_type)`), so this path can silently drop an existing `DEFAULT`/`MATERIALIZED`/`ALIAS...
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/Storages/AlterCommands.cpp b/src/Storages/AlterCommands.cpp index 3863adc67515..387e0f5e55c2 100644 --- a/src/Storages/AlterCommands.cpp +++ b/src/Storages/AlterCommands.cpp @@ -81,6 +81,7 @@ namespace ErrorCodes extern const int NOT_IMPLEMENTED; extern const int ALTER_OF_COLUMN_IS_FORBIDDEN;...
[ "src/Storages/AlterCommands.cpp" ]
[ { "comment": "When `ADD ENUM VALUES` sets `command.data_type` here, `default_kind`/`default_expression` are still unset for this command. Later, `AlterCommand::apply` treats any `data_type` change as a default rewrite (`if (default_expression || data_type)`), so this path can silently drop an existing `DEFAULT`...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
### Assertion doesn't guard against type with ADD ENUM VALUES **Low Severity** <!-- DESCRIPTION START --> The `chassert` verifying that column type isn't populated alongside special modifiers checks `command->remove_property`, `command_settings_changes`, and `command_settings_resets`, but omits `command_add_enum_valu...
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/Parsers/ParserAlterQuery.cpp b/src/Parsers/ParserAlterQuery.cpp index 866ce4286351..e2bd193b1a2d 100644 --- a/src/Parsers/ParserAlterQuery.cpp +++ b/src/Parsers/ParserAlterQuery.cpp @@ -45,6 +45,7 @@ bool ParserAlterCommand::parseImpl(Pos & pos, ASTPtr & node, Expected & expected ParserKeyword s_m...
[ "src/Parsers/ParserAlterQuery.cpp" ]
[ { "comment": "### Assertion doesn't guard against type with ADD ENUM VALUES\n\n**Low Severity**\n\n<!-- DESCRIPTION START -->\nThe `chassert` verifying that column type isn't populated alongside special modifiers checks `command->remove_property`, `command_settings_changes`, and `command_settings_resets`, but o...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
⚠️ `DataTypeEnum.h` is a high-fanout header, and adding a direct include of `Parsers/IAST.h` here increases transitive compile cost across many translation units. Please switch this to a forward declaration include (for example `Parsers/IAST_fwd.h`) and keep the full parser header out of this core type header.
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/DataTypes/DataTypeEnum.h b/src/DataTypes/DataTypeEnum.h index 86baf4238f9a..8e5222499559 100644 --- a/src/DataTypes/DataTypeEnum.h +++ b/src/DataTypes/DataTypeEnum.h @@ -1,8 +1,10 @@ #pragma once +#include <vector> #include <DataTypes/IDataType.h> #include <DataTypes/EnumValues.h> #include <Colum...
[ "src/DataTypes/DataTypeEnum.h" ]
[ { "comment": "⚠️ `DataTypeEnum.h` is a high-fanout header, and adding a direct include of `Parsers/IAST.h` here increases transitive compile cost across many translation units. Please switch this to a forward declaration include (for example `Parsers/IAST_fwd.h`) and keep the full parser header out of this core...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
πŸ’‘ Typo in SQL syntax example: `ADD ENUM VALUES (EnumName [= number] , ...]);` has an extra `]` before `);`. It should be `ADD ENUM VALUES (EnumName [= number], ...);`.
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/docs/en/sql-reference/statements/alter/column.md b/docs/en/sql-reference/statements/alter/column.md index f845210447bc..39f8fa02bb7c 100644 --- a/docs/en/sql-reference/statements/alter/column.md +++ b/docs/en/sql-reference/statements/alter/column.md @@ -29,6 +29,7 @@ The following actions are supported: -...
[ "docs/en/sql-reference/statements/alter/column.md" ]
[ { "comment": "πŸ’‘ Typo in SQL syntax example: `ADD ENUM VALUES (EnumName [= number] , ...]);` has an extra `]` before `);`. It should be `ADD ENUM VALUES (EnumName [= number], ...);`.", "path": "docs/en/sql-reference/statements/alter/column.md", "hunk": "@@ -275,6 +282,24 @@ Reset column setting `max_com...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
### Duplicate default expression preservation block after enum merge **Low Severity** <!-- DESCRIPTION START --> The block at lines 1512–1516 (`if (command.data_type && !command.default_expression && ...)`) is an exact duplicate of the block at lines 1435–1439. The second copy was added to handle the case where `comm...
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/Storages/AlterCommands.cpp b/src/Storages/AlterCommands.cpp index 3863adc67515..387e0f5e55c2 100644 --- a/src/Storages/AlterCommands.cpp +++ b/src/Storages/AlterCommands.cpp @@ -81,6 +81,7 @@ namespace ErrorCodes extern const int NOT_IMPLEMENTED; extern const int ALTER_OF_COLUMN_IS_FORBIDDEN;...
[ "src/Storages/AlterCommands.cpp" ]
[ { "comment": "### Duplicate default expression preservation block after enum merge\n\n**Low Severity**\n\n<!-- DESCRIPTION START -->\nThe block at lines 1512–1516 (`if (command.data_type && !command.default_expression && ...)`) is an exact duplicate of the block at lines 1435–1439. The second copy was added to ...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
⚠️ `ADD ENUM VALUES` validates the enum payload before validating that the target column is actually `Enum`. In this block, `tryCreateAddToEnumType` runs before checking `eti.enum_type`, so for non-enum columns an out-of-range literal (e.g. `1000`) throws `ARGUMENT_OUT_OF_BOUND` instead of the expected `ILLEGAL_COLUMN`...
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/Storages/AlterCommands.cpp b/src/Storages/AlterCommands.cpp index 3863adc67515..387e0f5e55c2 100644 --- a/src/Storages/AlterCommands.cpp +++ b/src/Storages/AlterCommands.cpp @@ -81,6 +81,7 @@ namespace ErrorCodes extern const int NOT_IMPLEMENTED; extern const int ALTER_OF_COLUMN_IS_FORBIDDEN;...
[ "src/Storages/AlterCommands.cpp" ]
[ { "comment": "⚠️ `ADD ENUM VALUES` validates the enum payload before validating that the target column is actually `Enum`. In this block, `tryCreateAddToEnumType` runs before checking `eti.enum_type`, so for non-enum columns an out-of-range literal (e.g. `1000`) throws `ARGUMENT_OUT_OF_BOUND` instead of the exp...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
`ADD ENUM VALUES` currently uses one global relative flag for the whole list. If any explicit assignment appears, earlier implicit entries are treated as absolute, so valid mixed forms can fail (for example, base max 2 with `ADD ENUM VALUES` using implicit first and explicit later maps the implicit value to 1 instead o...
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/DataTypes/DataTypeEnum.cpp b/src/DataTypes/DataTypeEnum.cpp index 682cb3106b26..e401538cfdc8 100644 --- a/src/DataTypes/DataTypeEnum.cpp +++ b/src/DataTypes/DataTypeEnum.cpp @@ -16,6 +16,7 @@ #include <Interpreters/Context.h> #include <Core/Settings.h> +#include <algorithm> #include <limits> @...
[ "src/DataTypes/DataTypeEnum.cpp" ]
[ { "comment": "`ADD ENUM VALUES` currently uses one global relative flag for the whole list. If any explicit assignment appears, earlier implicit entries are treated as absolute, so valid mixed forms can fail (for example, base max 2 with `ADD ENUM VALUES` using implicit first and explicit later maps the implici...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
⚠️ Test coverage misses an important error-priority case for `ADD ENUM VALUES` on non-enum columns. This test checks `ILLEGAL_COLUMN` only for an in-range value (`'A' = 1`). Please add a companion case with an out-of-range literal (for example `1000`) and still expect `ILLEGAL_COLUMN`. Without this, a regression can ...
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/tests/queries/0_stateless/04307_add_to_enum.sql.j2 b/tests/queries/0_stateless/04307_add_to_enum.sql.j2 new file mode 100644 index 000000000000..250dd7fdf109 --- /dev/null +++ b/tests/queries/0_stateless/04307_add_to_enum.sql.j2 @@ -0,0 +1,129 @@ +{% for NULLABLE_START, NULLABLE_END in {'' : '', 'Nullable(...
[ "tests/queries/0_stateless/04307_add_to_enum.sql.j2" ]
[ { "comment": "⚠️ Test coverage misses an important error-priority case for `ADD ENUM VALUES` on non-enum columns.\n\nThis test checks `ILLEGAL_COLUMN` only for an in-range value (`'A' = 1`). Please add a companion case with an out-of-range literal (for example `1000`) and still expect `ILLEGAL_COLUMN`.\n\nWitho...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
`autoAssignNumberForEnum` now rejects any explicit assignment after a leading implicit prefix when the explicit value is `<= relative_max`. That introduces a backward-compatibility regression in general `Enum` parsing (not only `ADD ENUM VALUES`): declarations with intentionally absolute explicit values (for example n...
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/DataTypes/DataTypeEnum.cpp b/src/DataTypes/DataTypeEnum.cpp index 682cb3106b26..e401538cfdc8 100644 --- a/src/DataTypes/DataTypeEnum.cpp +++ b/src/DataTypes/DataTypeEnum.cpp @@ -16,6 +16,7 @@ #include <Interpreters/Context.h> #include <Core/Settings.h> +#include <algorithm> #include <limits> @...
[ "src/DataTypes/DataTypeEnum.cpp" ]
[ { "comment": "`autoAssignNumberForEnum` now rejects any explicit assignment after a leading implicit prefix when the explicit value is `<= relative_max`.\n\nThat introduces a backward-compatibility regression in general `Enum` parsing (not only `ADD ENUM VALUES`): declarations with intentionally absolute explic...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
⚠️ Documentation and examples in this PR show that `ADD ENUM VALUES` also works for `Nullable(Enum*)`, but this sentence mentions only non-nullable enum types. Please mention `Nullable(Enum)`, `Nullable(Enum8)`, and `Nullable(Enum16)` to avoid user confusion.
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/docs/en/sql-reference/statements/alter/column.md b/docs/en/sql-reference/statements/alter/column.md index f845210447bc..39f8fa02bb7c 100644 --- a/docs/en/sql-reference/statements/alter/column.md +++ b/docs/en/sql-reference/statements/alter/column.md @@ -29,6 +29,7 @@ The following actions are supported: -...
[ "docs/en/sql-reference/statements/alter/column.md" ]
[ { "comment": "⚠️ Documentation and examples in this PR show that `ADD ENUM VALUES` also works for `Nullable(Enum*)`, but this sentence mentions only non-nullable enum types. Please mention `Nullable(Enum)`, `Nullable(Enum8)`, and `Nullable(Enum16)` to avoid user confusion.\n", "path": "docs/en/sql-reference...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
Small docs typo: `syntax sugar` is usually written as `syntactic sugar`.
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/docs/en/sql-reference/data-types/enum.md b/docs/en/sql-reference/data-types/enum.md index 5f330bcd8766..ecbc90e8bc42 100644 --- a/docs/en/sql-reference/data-types/enum.md +++ b/docs/en/sql-reference/data-types/enum.md @@ -162,3 +162,26 @@ Enum values are also convertible to numeric types using the `toT` fu...
[ "docs/en/sql-reference/data-types/enum.md" ]
[ { "comment": "Small docs typo: `syntax sugar` is usually written as `syntactic sugar`.\n", "path": "docs/en/sql-reference/data-types/enum.md", "hunk": "@@ -164,3 +164,26 @@ Enum values are also convertible to numeric types using the `toT` function, wher\n The Enum type can be changed without cost using ...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
```suggestion factory.registerDataType("Enum16", [](const ASTPtr & arguments) { return createExact<DataTypeEnum16>(arguments, false); }); ```
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/DataTypes/DataTypeEnum.cpp b/src/DataTypes/DataTypeEnum.cpp index 682cb3106b26..e401538cfdc8 100644 --- a/src/DataTypes/DataTypeEnum.cpp +++ b/src/DataTypes/DataTypeEnum.cpp @@ -16,6 +16,7 @@ #include <Interpreters/Context.h> #include <Core/Settings.h> +#include <algorithm> #include <limits> @...
[ "src/DataTypes/DataTypeEnum.cpp" ]
[ { "comment": "```suggestion\n factory.registerDataType(\"Enum16\", [](const ASTPtr & arguments)\n {\n return createExact<DataTypeEnum16>(arguments, false);\n });\n```", "path": "src/DataTypes/DataTypeEnum.cpp", "hunk": "@@ -332,17 +362,88 @@ static DataTypePtr create(const ASTPtr & argum...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
```suggestion factory.registerDataType("Enum", [](const ASTPtr & arguments) { return create(arguments, false); }); ```
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/DataTypes/DataTypeEnum.cpp b/src/DataTypes/DataTypeEnum.cpp index 682cb3106b26..e401538cfdc8 100644 --- a/src/DataTypes/DataTypeEnum.cpp +++ b/src/DataTypes/DataTypeEnum.cpp @@ -16,6 +16,7 @@ #include <Interpreters/Context.h> #include <Core/Settings.h> +#include <algorithm> #include <limits> @...
[ "src/DataTypes/DataTypeEnum.cpp" ]
[ { "comment": "```suggestion\n factory.registerDataType(\"Enum\", [](const ASTPtr & arguments)\n {\n return create(arguments, false);\n });\n```", "path": "src/DataTypes/DataTypeEnum.cpp", "hunk": "@@ -332,17 +362,88 @@ static DataTypePtr create(const ASTPtr & arguments)\n Int64 v...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
```suggestion ```
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/Storages/AlterCommands.cpp b/src/Storages/AlterCommands.cpp index 3863adc67515..387e0f5e55c2 100644 --- a/src/Storages/AlterCommands.cpp +++ b/src/Storages/AlterCommands.cpp @@ -81,6 +81,7 @@ namespace ErrorCodes extern const int NOT_IMPLEMENTED; extern const int ALTER_OF_COLUMN_IS_FORBIDDEN;...
[ "src/Storages/AlterCommands.cpp" ]
[ { "comment": "```suggestion\n```", "path": "src/Storages/AlterCommands.cpp", "hunk": "@@ -1437,12 +1447,82 @@ void AlterCommands::prepare(const StorageInMemoryMetadata & metadata, bool share\n if (has_column)\n {\n const auto & column_from_table = columns.get(comm...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
Small docs typo in the syntax snippet: there is an extra space before the comma in ``'EnumName' [= number] , ...``. Could you change it to ``'EnumName' [= number], ...`` for consistency with the rest of SQL syntax docs?
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/docs/en/sql-reference/statements/alter/column.md b/docs/en/sql-reference/statements/alter/column.md index f845210447bc..39f8fa02bb7c 100644 --- a/docs/en/sql-reference/statements/alter/column.md +++ b/docs/en/sql-reference/statements/alter/column.md @@ -29,6 +29,7 @@ The following actions are supported: -...
[ "docs/en/sql-reference/statements/alter/column.md" ]
[ { "comment": "Small docs typo in the syntax snippet: there is an extra space before the comma in ``'EnumName' [= number] , ...``.\n\nCould you change it to ``'EnumName' [= number], ...`` for consistency with the rest of SQL syntax docs?\n", "path": "docs/en/sql-reference/statements/alter/column.md", "hu...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
`prepare` merges this `ADD ENUM VALUES` payload against `column_from_table`, but `column_from_table` comes from the original `columns` snapshot and that snapshot is never advanced after earlier commands in the same `ALTER`. For example, `ALTER TABLE t MODIFY COLUMN x ADD ENUM VALUES('a'), MODIFY COLUMN x ADD ENUM VALUE...
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/Storages/AlterCommands.cpp b/src/Storages/AlterCommands.cpp index 3863adc67515..387e0f5e55c2 100644 --- a/src/Storages/AlterCommands.cpp +++ b/src/Storages/AlterCommands.cpp @@ -81,6 +81,7 @@ namespace ErrorCodes extern const int NOT_IMPLEMENTED; extern const int ALTER_OF_COLUMN_IS_FORBIDDEN;...
[ "src/Storages/AlterCommands.cpp" ]
[ { "comment": "`prepare` merges this `ADD ENUM VALUES` payload against `column_from_table`, but `column_from_table` comes from the original `columns` snapshot and that snapshot is never advanced after earlier commands in the same `ALTER`. For example, `ALTER TABLE t MODIFY COLUMN x ADD ENUM VALUES('a'), MODIFY C...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
`ADD ENUM VALUES` still derives from a stale snapshot when it follows a normal type `MODIFY COLUMN` in the same `ALTER`. This line advances `columns` only after an `ADD ENUM VALUES`; a preceding command like `MODIFY COLUMN x Enum8('a' = 1, 'b' = 2), MODIFY COLUMN x ADD ENUM VALUES('c')` leaves `columns` at the original...
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/Storages/AlterCommands.cpp b/src/Storages/AlterCommands.cpp index 3863adc67515..387e0f5e55c2 100644 --- a/src/Storages/AlterCommands.cpp +++ b/src/Storages/AlterCommands.cpp @@ -81,6 +81,7 @@ namespace ErrorCodes extern const int NOT_IMPLEMENTED; extern const int ALTER_OF_COLUMN_IS_FORBIDDEN;...
[ "src/Storages/AlterCommands.cpp" ]
[ { "comment": "`ADD ENUM VALUES` still derives from a stale snapshot when it follows a normal type `MODIFY COLUMN` in the same `ALTER`. This line advances `columns` only after an `ADD ENUM VALUES`; a preceding command like `MODIFY COLUMN x Enum8('a' = 1, 'b' = 2), MODIFY COLUMN x ADD ENUM VALUES('c')` leaves `co...
true
ClickHouse/ClickHouse
93,830
comment_to_fix
Support for ALTER TABLE ADD ENUM VALUES
This guard is global for the whole `ALTER`, so it rejects independent column changes that should not interact. For example, `ALTER TABLE t MODIFY COLUMN y UInt16, MODIFY COLUMN x ADD ENUM VALUES('b')` now throws `NOT_IMPLEMENTED` just because a previous command modified any column type, even though the stale-snapshot p...
8a69decd78adb17700abfad4538c5d3750226194
07520f54fbc7a36adc637c461574cfc15b3a8c8e
diff --git a/src/Storages/AlterCommands.cpp b/src/Storages/AlterCommands.cpp index 3863adc67515..387e0f5e55c2 100644 --- a/src/Storages/AlterCommands.cpp +++ b/src/Storages/AlterCommands.cpp @@ -81,6 +81,7 @@ namespace ErrorCodes extern const int NOT_IMPLEMENTED; extern const int ALTER_OF_COLUMN_IS_FORBIDDEN;...
[ "src/Storages/AlterCommands.cpp" ]
[ { "comment": "This guard is global for the whole `ALTER`, so it rejects independent column changes that should not interact. For example, `ALTER TABLE t MODIFY COLUMN y UInt16, MODIFY COLUMN x ADD ENUM VALUES('b')` now throws `NOT_IMPLEMENTED` just because a previous command modified any column type, even thoug...
true
ClickHouse/ClickHouse
58,957
issue_to_patch
Parse detached part name properly in case of `tryN` suffix
### Changelog category (leave one): - Bug Fix (user-visible misbehavior in an official stable release) ### Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md): Make possible to drop detached parts with the `tryN` suffix. ### Documentation entry for user-facing changes ...
b06c9fc1e8d0589a2689a964f841dfd2c4dde43b
5a97c47f8cc44684f949f702e24274d41258eb00
diff --git a/src/Storages/MergeTree/DataPartStorageOnDiskBase.cpp b/src/Storages/MergeTree/DataPartStorageOnDiskBase.cpp index c4e6846a48d1..03d3c603fb41 100644 --- a/src/Storages/MergeTree/DataPartStorageOnDiskBase.cpp +++ b/src/Storages/MergeTree/DataPartStorageOnDiskBase.cpp @@ -146,7 +146,7 @@ String DataPartStorag...
[ "src/Storages/MergeTree/DataPartStorageOnDiskBase.cpp", "src/Storages/MergeTree/MergeTreeData.cpp", "src/Storages/MergeTree/MergeTreePartInfo.cpp", "src/Storages/MergeTree/MergeTreePartInfo.h", "src/Storages/StorageReplicatedMergeTree.cpp", "tests/queries/0_stateless/04063_drop_detached_part_with_try_n_su...
[ { "comment": "```suggestion\n chassert(dir_name.size() >= 2);\n```", "path": "src/Storages/MergeTree/MergeTreePartInfo.cpp", "hunk": "@@ -343,7 +346,18 @@ DetachedPartInfo DetachedPartInfo::parseDetachedPartName(\n part_info.disk = disk;\n part_info.dir_name = dir_name;\n \n- /// First...
diff --git a/tests/queries/0_stateless/04063_drop_detached_part_with_try_n_suffix.reference b/tests/queries/0_stateless/04063_drop_detached_part_with_try_n_suffix.reference new file mode 100644 index 000000000000..8accb1745c93 --- /dev/null +++ b/tests/queries/0_stateless/04063_drop_detached_part_with_try_n_suffix.refe...
true
ClickHouse/ClickHouse
58,957
comment_to_fix
Parse detached part name properly in case of `tryN` suffix
```suggestion chassert(dir_name.size() >= 2); ```
b06c9fc1e8d0589a2689a964f841dfd2c4dde43b
5a97c47f8cc44684f949f702e24274d41258eb00
diff --git a/src/Storages/MergeTree/MergeTreePartInfo.cpp b/src/Storages/MergeTree/MergeTreePartInfo.cpp index 346b4bb09d75..c41ab2b4fcc1 100644 --- a/src/Storages/MergeTree/MergeTreePartInfo.cpp +++ b/src/Storages/MergeTree/MergeTreePartInfo.cpp @@ -1,9 +1,12 @@ #include <Storages/MergeTree/MergeTreePartInfo.h> + +#i...
[ "src/Storages/MergeTree/MergeTreePartInfo.cpp" ]
[ { "comment": "```suggestion\n chassert(dir_name.size() >= 2);\n```", "path": "src/Storages/MergeTree/MergeTreePartInfo.cpp", "hunk": "@@ -343,7 +346,18 @@ DetachedPartInfo DetachedPartInfo::parseDetachedPartName(\n part_info.disk = disk;\n part_info.dir_name = dir_name;\n \n- /// First...
true
ClickHouse/ClickHouse
58,957
comment_to_fix
Parse detached part name properly in case of `tryN` suffix
`parseDetachedPartName` currently strips `"_try"` even when there are no trailing digits (e.g. `all_0_0_0_try`), because `std::all_of` on an empty range returns `true`. It also only accepts up to two digits, while the suffix name is `tryN`. This can misclassify malformed detached directory names as valid and still mis...
b06c9fc1e8d0589a2689a964f841dfd2c4dde43b
5a97c47f8cc44684f949f702e24274d41258eb00
diff --git a/src/Storages/MergeTree/MergeTreePartInfo.cpp b/src/Storages/MergeTree/MergeTreePartInfo.cpp index 346b4bb09d75..c41ab2b4fcc1 100644 --- a/src/Storages/MergeTree/MergeTreePartInfo.cpp +++ b/src/Storages/MergeTree/MergeTreePartInfo.cpp @@ -1,9 +1,12 @@ #include <Storages/MergeTree/MergeTreePartInfo.h> + +#i...
[ "src/Storages/MergeTree/MergeTreePartInfo.cpp" ]
[ { "comment": "`parseDetachedPartName` currently strips `\"_try\"` even when there are no trailing digits (e.g. `all_0_0_0_try`), because `std::all_of` on an empty range returns `true`. It also only accepts up to two digits, while the suffix name is `tryN`.\n\nThis can misclassify malformed detached directory na...
true
ClickHouse/ClickHouse
58,957
comment_to_fix
Parse detached part name properly in case of `tryN` suffix
Please extend this regression test with boundary cases for the new suffix parser behavior: - a malformed name without trailing digits (for example `all_0_0_0_try`) should stay invalid and not be treated as `tryN`; - a multi-digit suffix (for example `all_0_0_0_try100`) should be accepted and droppable. Without these ...
b06c9fc1e8d0589a2689a964f841dfd2c4dde43b
5a97c47f8cc44684f949f702e24274d41258eb00
diff --git a/tests/queries/0_stateless/04063_drop_detached_part_with_try_n_suffix.sh b/tests/queries/0_stateless/04063_drop_detached_part_with_try_n_suffix.sh new file mode 100755 index 000000000000..d233b70f1e30 --- /dev/null +++ b/tests/queries/0_stateless/04063_drop_detached_part_with_try_n_suffix.sh @@ -0,0 +1,130 ...
[ "tests/queries/0_stateless/04063_drop_detached_part_with_try_n_suffix.sh" ]
[ { "comment": "Please extend this regression test with boundary cases for the new suffix parser behavior:\n\n- a malformed name without trailing digits (for example `all_0_0_0_try`) should stay invalid and not be treated as `tryN`;\n- a multi-digit suffix (for example `all_0_0_0_try100`) should be accepted and d...
true
ClickHouse/ClickHouse
58,957
comment_to_fix
Parse detached part name properly in case of `tryN` suffix
This makes `all_0_0_0_try1` a valid detached part with an empty `prefix`, but `ATTACH PARTITION ALL` still passes `part_info.dir_name` to `ActiveDataPartSet::tryAdd` / `getContainingPart` in `MergeTreeData::tryLoadPartsToAttach`. Those methods call `MergeTreePartInfo::fromPartName`, so the original suffixed directory n...
b06c9fc1e8d0589a2689a964f841dfd2c4dde43b
5a97c47f8cc44684f949f702e24274d41258eb00
diff --git a/src/Storages/MergeTree/MergeTreePartInfo.cpp b/src/Storages/MergeTree/MergeTreePartInfo.cpp index 346b4bb09d75..c41ab2b4fcc1 100644 --- a/src/Storages/MergeTree/MergeTreePartInfo.cpp +++ b/src/Storages/MergeTree/MergeTreePartInfo.cpp @@ -1,9 +1,12 @@ #include <Storages/MergeTree/MergeTreePartInfo.h> + +#i...
[ "src/Storages/MergeTree/MergeTreePartInfo.cpp" ]
[ { "comment": "This makes `all_0_0_0_try1` a valid detached part with an empty `prefix`, but `ATTACH PARTITION ALL` still passes `part_info.dir_name` to `ActiveDataPartSet::tryAdd` / `getContainingPart` in `MergeTreeData::tryLoadPartsToAttach`. Those methods call `MergeTreePartInfo::fromPartName`, so the origina...
true
ClickHouse/ClickHouse
58,957
comment_to_fix
Parse detached part name properly in case of `tryN` suffix
`has_try_suffix` is only checked in `MergeTreeData::tryLoadPartsToAttach`, but `StorageReplicatedMergeTree::attachPartHelperFoundValidPart` has a separate detached-part candidate scan for `ATTACH_PART` log entries and still keeps any part with `valid_name && prefix.empty`. With this parser, a leftover `all_0_0_0_try1` ...
b06c9fc1e8d0589a2689a964f841dfd2c4dde43b
5a97c47f8cc44684f949f702e24274d41258eb00
diff --git a/src/Storages/MergeTree/MergeTreePartInfo.h b/src/Storages/MergeTree/MergeTreePartInfo.h index 599093d820cc..f1ca8c0e2529 100644 --- a/src/Storages/MergeTree/MergeTreePartInfo.h +++ b/src/Storages/MergeTree/MergeTreePartInfo.h @@ -183,6 +183,12 @@ struct DetachedPartInfo : public MergeTreePartInfo /// ...
[ "src/Storages/MergeTree/MergeTreePartInfo.h" ]
[ { "comment": "`has_try_suffix` is only checked in `MergeTreeData::tryLoadPartsToAttach`, but `StorageReplicatedMergeTree::attachPartHelperFoundValidPart` has a separate detached-part candidate scan for `ATTACH_PART` log entries and still keeps any part with `valid_name && prefix.empty`. With this parser, a left...
true
ClickHouse/ClickHouse
97,014
issue_to_patch
Fix bugfix validation failing due to git ownership check in Docker
## Summary - The `find_master_builds` function (introduced in #96868) runs `git log origin/master` inside Docker containers that run as root, but the repository is owned by the `ubuntu` user - Git's ownership check blocks the command, returning empty output, so no master commits are found and the job crashes with `asse...
d0534d76c0718cf7d990e929bf22660641ae4a90
75f70593f94f80f92b7be4250fa0de20ce47d032
diff --git a/.github/workflows/pull_request.yml b/.github/workflows/pull_request.yml index 4133e1e09402..980b5711112e 100644 --- a/.github/workflows/pull_request.yml +++ b/.github/workflows/pull_request.yml @@ -1698,7 +1698,7 @@ jobs: PYTHONUNBUFFERED=1 python3 -m praktika run 'Integration tests (amd_asan_ub...
[ ".github/workflows/pull_request.yml", "ci/defs/job_configs.py", "ci/jobs/functional_tests.py", "ci/jobs/integration_test_job.py", "ci/jobs/scripts/bugfix_validation.py", "ci/jobs/scripts/clickhouse_proc.py", "ci/jobs/scripts/workflow_hooks/store_data.py" ]
[ { "comment": "this job stage checks crashes which are, in context of bugfix validation, \"success\" case (bug reproduced), the stage supposed to be executed for every build for job to work correctly\n", "path": "ci/jobs/functional_tests.py", "hunk": "", "resolving_sha": "75f70593f94f80f92b7be4250fa0...
diff --git a/ci/jobs/integration_test_job.py b/ci/jobs/integration_test_job.py index e10e26c92b5f..1afe89c6a602 100644 --- a/ci/jobs/integration_test_job.py +++ b/ci/jobs/integration_test_job.py @@ -8,6 +8,7 @@ from more_itertools import tail +from ci.jobs.scripts.bugfix_validation import BUGFIX_BUILD_TYPES, find_...
true
ClickHouse/ClickHouse
97,014
comment_to_fix
Fix bugfix validation failing due to git ownership check in Docker
this job stage checks crashes which are, in context of bugfix validation, "success" case (bug reproduced), the stage supposed to be executed for every build for job to work correctly
d0534d76c0718cf7d990e929bf22660641ae4a90
75f70593f94f80f92b7be4250fa0de20ce47d032
diff --git a/ci/jobs/functional_tests.py b/ci/jobs/functional_tests.py index a6b7388b8ce8..345fe10eaf8d 100644 --- a/ci/jobs/functional_tests.py +++ b/ci/jobs/functional_tests.py @@ -5,6 +5,7 @@ import subprocess from pathlib import Path +from ci.jobs.scripts.bugfix_validation import BUGFIX_BUILD_TYPES, find_master...
[ "ci/jobs/functional_tests.py" ]
[ { "comment": "this job stage checks crashes which are, in context of bugfix validation, \"success\" case (bug reproduced), the stage supposed to be executed for every build for job to work correctly\n", "path": "ci/jobs/functional_tests.py", "hunk": "", "resolving_sha": "75f70593f94f80f92b7be4250fa0...
true