[CI] Refactor toolchain-artifact related inputs/outputs #19643
Conversation
* Group all artifact-related inputs in `sycl-linux-build.yml`. I plan on
refactoring artifacts for E2E tests/release toolchain in a subsequent PR.
* Unify inputs/outputs names between `sycl-*-build.yml` and
`sycl-*-run-tests.yml`. Windows still has less flexibility, but for the
inputs that exist there the names still match Linux and the inputs are
ordered similarly to Linux.
* Add `${{ needs.build.outputs.toolchain_artifact }}` instead of
duplicating the artifact name. Aligns usage with actual
filename/decompress command.
aelovikov-intel
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
aelovikov-intel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
aelovikov-intel
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
| value: ${{ inputs.toolchain_artifact }} | ||
| toolchain_artifact_filename: | ||
| value: ${{ jobs.build.outputs.toolchain_artifact_filename }} | ||
| toolchain_decompress_command: |
There was a problem hiding this comment.
not directly related but do we actually need toolchain_decompress_command? If I remember correctly there were only like two cases we actually use, maybe it was tar.gz and tar.zst. If that's true could we get rid of this var everywhere and just have an if statement when we actually run the decompression to handle both cases? I don't think we need to support arbitrary decompression commands specified by higher level workflows
Maybe we could also just pass the filename blindly to 7z or something which probably figures out what to do internally without us having to have conditionals.
There was a problem hiding this comment.
Yes, we only have two. However, doing the if at the use-case would mean that every user (well, it's only sycl-hardening-check.yml and sycl-linux-run-tests.yml for now, but still) will have to duplicate it, which I'd dislike even more than this.
Ideally, we could have a single (JSON?) output that contained all the data. That is something I'm leaving for future/somebody else :)
There was a problem hiding this comment.
to be honest i think we could just slap the files to 7z but yeah we can leave it to someone else
| @@ -281,7 +288,7 @@ jobs: | |||
| if: ${{ always() && !cancelled() && steps.build.conclusion == 'success' }} | |||
| uses: actions/upload-artifact@v4 | |||
| with: | |||
| name: sycl_linux_${{ inputs.build_artifact_suffix }} | |||
There was a problem hiding this comment.
lol why did we ever do this
aelovikov-intel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
aelovikov-intel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
aelovikov-intel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
aelovikov-intel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
aelovikov-intel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
aelovikov-intel
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
|
Failures in Pre/Post-commit tasks are on individual tests and not due to a regression in CI itself, unrelated to this PR. Nightly run: https://github.com/intel/llvm/actions/runs/16630331081 was a separate regression and has been addressed in #19641 Benchmarks workflow: https://github.com/intel/llvm/actions/runs/16630629189 |
sycl-linux-build.yml. I plan on refactoring artifacts for E2E tests/release toolchain in a subsequent PR.sycl-*-build.ymlandsycl-*-run-tests.yml. Windows still has less flexibility, but for the inputs that exist there the names still match Linux and the inputs are ordered similarly to Linux.${{ needs.build.outputs.toolchain_artifact }}instead of duplicating the artifact name. Aligns usage with actual filename/decompress command.