[libc][workflow] improve ci coverage with windows-2025 and arm ubuntu #123745
Conversation
|
@llvm/pr-subscribers-github-workflow Author: Schrodinger ZHU Yifan (SchrodingerZhu) ChangesAdd the following workflows:
Full diff: https://github.com/llvm/llvm-project/pull/123745.diff 2 Files Affected:
diff --git a/.github/workflows/libc-fullbuild-tests.yml b/.github/workflows/libc-fullbuild-tests.yml
index 58e15ce29546ef..1db69665ae186d 100644
--- a/.github/workflows/libc-fullbuild-tests.yml
+++ b/.github/workflows/libc-fullbuild-tests.yml
@@ -11,12 +11,16 @@ on:
jobs:
build:
- runs-on: ubuntu-24.04
+ runs-on: ${{ matrix.os }}
strategy:
fail-fast: false
matrix:
include:
- - c_compiler: clang
+ - os: ubuntu-24.04
+ c_compiler: clang
+ cpp_compiler: clang++
+ - os: ubuntu-24.04-arm
+ c_compiler: clang
cpp_compiler: clang++
# TODO: add back gcc build when it is fixed
# - c_compiler: gcc
diff --git a/.github/workflows/libc-overlay-tests.yml b/.github/workflows/libc-overlay-tests.yml
index 8b59d76aed4a88..62c99742f7a6cc 100644
--- a/.github/workflows/libc-overlay-tests.yml
+++ b/.github/workflows/libc-overlay-tests.yml
@@ -22,10 +22,18 @@ jobs:
compiler:
c_compiler: clang
cpp_compiler: clang++
+ - os: ubuntu-24.04-arm
+ compiler:
+ c_compiler: clang
+ cpp_compiler: clang++
- os: windows-2022
compiler:
c_compiler: clang-cl
cpp_compiler: clang-cl
+ - os: windows-2025
+ compiler:
+ c_compiler: clang-cl
+ cpp_compiler: clang-cl
- os: macos-14
compiler:
c_compiler: clang
|
SchrodingerZhu
force-pushed
the
libc-workflow-win2025-linuxarm
branch
from
2b2bc84 to
d26214b
Compare
SchrodingerZhu
requested review from
statham-arm,
nickdesaulniers,
lntue,
michaelrj-google and
jhuber6
|
Can you take a look at (macos-14, clang, clang++) failures? |
SchrodingerZhu
changed the title
|
Switched to |
|
@SchrodingerZhu Looks good in general. I don't know much about libc, though. |
|
@tstellar thanks! I will also be waiting for @lntue, @boomanaiden154 and others' feedback then. |
| strategy: | ||
| fail-fast: false | ||
| matrix: | ||
| include: | ||
| - c_compiler: clang | ||
| - os: ubuntu-24.04 | ||
| ccache-variant: sccache |
There was a problem hiding this comment.
Why use sccache for this, but ccache for the arm variant?
There was a problem hiding this comment.
sccache is a modern implementation from Mozilla, it supports more storage format and compiler (e.g. msvc/clang-cl). I prefer to default on sccache but due to hendrikmuhs/ccache-action#279, ccache is used on arm64 ubuntu.
There was a problem hiding this comment.
Please add a comment linking to that issue in the yml file. I assume it will be resolved some day, so if someone notices that in the future, it would be good to then resolve the arm specific differences.
There was a problem hiding this comment.
Seems reasonable enough to me.
We need to be a bit careful with how many jobs run at once, but from what I've seen the libc jobs are reasonably fast, so shouldn't be a big issue. I'm not too sure how the new ARM runners counter against the limits though (their own lower limit? the general linux limit?).
Once the new premerge is spun up, I would like to reevaluate what configurations we're testing here too. We definitely can't test all the new configurations in the monorepo-wide premerge pipeline, but we do want to test a "standard" build on Linux (and windows if reasonable), and it would be good to not duplicate that with what's being done here.
|
Merge this according to previous approval. New updates are trivial TODO lines. |
Add the following workflows:
fullbuildon aarch64 ubuntuoverlayon windows 2025overlayon aarch64 ubuntuccachevariant is used onaarch64due to hendrikmuhs/ccache-action#279