Repository navigation
Conversation
The input only reaches family selection on workflow_dispatch. Pull requests start from TheRock's presubmit defaults and filter by host-asan variant support, so the default here does not narrow PR scope. Co-authored-by: Cursor <cursoragent@cursor.com>
…ults Sanity currently fails on a LeakSanitizer finding in libamdhip64, which skips the two component test jobs this workflow exists to run. Set sanity_blocks_components so they run anyway; sanity still reports its own failure, so the leak stays visible while it is investigated. Co-authored-by: Cursor <cursoragent@cursor.com>
Point to users/raramakr/asan-presubmit-test branch which enables host-asan tests for external repos with build_stages set. Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>
TheRock now keys host-asan sandbox test runners off an explicit presubmit input rather than inferring intent from build_stages being set. Opt in so the runtime-only presubmit keeps getting a test runner on pull requests. Co-authored-by: Cursor <cursoragent@cursor.com>
Follows the rename on TheRock's users/raramakr/asan-presubmit-test branch, which this workflow pins by name. Co-authored-by: Cursor <cursoragent@cursor.com>
Resolves the same #11233 conflict as #10558: develop disabled Python/PyTorch/JAX for the ASAN workflow while this branch was already doing so as part of scoping the host-ASAN presubmit. Keeps develop's rationale comment and this branch's single copy of the three inputs, since taking both sides would produce duplicate YAML keys. Co-authored-by: Cursor <cursoragent@cursor.com>
Validate ROCm/TheRock#7780 on top of current main (which already carries ROCm/TheRock#7602 and ROCm/TheRock#7600) without the sanity-check fixes still in review as ROCm/TheRock#8100. Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
…n_test Co-authored-by: Cursor <cursoragent@cursor.com> # Conflicts: # .github/workflows/therock-multi-arch-ci-asan.yml
Co-authored-by: Cursor <cursoragent@cursor.com>
…output) Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
…ranching) Co-authored-by: Cursor <cursoragent@cursor.com>
…ck preload Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
…n_test Co-authored-by: Cursor <cursoragent@cursor.com> # Conflicts: # .github/workflows/therock-multi-arch-ci-asan.yml
Co-authored-by: Cursor <cursoragent@cursor.com>
…n_test Co-authored-by: Cursor <cursoragent@cursor.com> # Conflicts: # .github/workflows/therock-multi-arch-ci-asan.yml
Points the host-asan presubmit at users/dravindr/host_asan_8455_test, which carries ROCm/TheRock#8455 (test_triggers as data in all_build_variants) in place of ROCm/TheRock#7780 (the asan_presubmit workflow input). #7780 enabled presubmit tests only for external_repo callers that opted in. #8455 has no such condition: host-asan lists pull_request in its test_triggers, so any host-asan presubmit build runs its tests. The input is therefore gone from this workflow, and the run should otherwise look like the last one. Not for merge -- validation branch. Co-authored-by: Cursor <cursoragent@cursor.com>
github-actions
Bot
added
the
github actions
✅ All Policy Checks Passed
📖 Need help? See the Policy FAQ for details on every check and how to fix failures. |
…n_test_8455 Co-authored-by: Cursor <cursoragent@cursor.com> # Conflicts: # .github/workflows/therock-multi-arch-ci-asan.yml
users/dravindr/host_asan_8455_test now carries the post-review #8455 (tests_on_trigger + _get_current_triggers) instead of the original event-name version. Empty commit to re-trigger. Co-authored-by: Cursor <cursoragent@cursor.com>
Keeps the TheRock refs pointed at users/dravindr/host_asan_8455_test (which carries TheRock#8455 and TheRock#8341) rather than develop's pinned SHA, and refreshes a stale comment that still referred to the pre-review test_triggers/pull_request vocabulary.
The develop merge took our side of the job block, which reverted multi_arch_ci_asan_summary back to the older multi_arch_ci_summary.
TheRock#8800 added path-based skip filtering for external repos. With no pr_number or SHAs in external_repo it takes the legacy branch and runs `git diff --name-only HEAD^` inside the shallow external-repo checkout, which exits 128 and fails setup.
TheRock#8800 skips CI for external repos when every changed file is skippable. The workflow yml alone is skippable, so this run needs one non-skippable file. Comment-only; revert before merge.
TheRock host_asan_8455_test now starts hip-tests and rocrtst even when the sanity check fails. This push retriggers that validation run. Co-authored-by: Cursor <cursoragent@cursor.com>
Keep the host-asan test-branch refs so the non-blocking sanity run can start, and take develop's external_repo fields for the path filter. Co-authored-by: Cursor <cursoragent@cursor.com>
cfreeamd
reviewed
Oct 20, 2026
cfreeamd
left a comment
cfreeamd
left a comment
Contributor
There was a problem hiding this comment.
🤖 AI-assisted pre-review (for @cfreeamd)
Two CI-config files; no new test logic added to os.cc (the change is purely a trailing comment to force a rebuild). The rocrtst_yaml_missing flag is a false positive — no new TEST(...) entries were added. The malicious_intent_risk flag (workflow change) is expected for this kind of CI validation PR; no supply-chain concern.
Before any merge: (1) replace the three mutable ref: users/dravindr/host_asan_8455_test branch pins with pinned commit SHAs, and (2) revert the os.cc comment touch (the comment already says this — just a reminder).
| @@ -57,20 +65,30 @@ concurrency: | |||
|
|
|||
| jobs: | |||
Contributor
There was a problem hiding this comment.
🤖 AI-assisted pre-review (for @cfreeamd)
[Required] This pins to a mutable branch ref (users/dravindr/host_asan_8455_test) rather than a commit SHA. The original lines used c01e3a3f3f67966dd35612cf363b49964f33cd47 — branch refs can be force-pushed, silently changing what the CI runs. All three ref: occurrences here must be replaced with a pinned SHA before merge.
|
|
||
| } // namespace rocrtst | ||
|
|
||
| // Touched to force a CI build for host-ASAN validation (TheRock#8455). Revert before merge. |
Contributor
There was a problem hiding this comment.
🤖 AI-assisted pre-review (for @cfreeamd)
[Required] This comment-touch must be reverted before merge (as the comment itself notes). No functional change — just a CI trigger artifact.
Trusting CMAKE_*_FLAGS_INIT left libamdhip64 linked with the ASan runtime and no __asan_init, so hipFree aborted on a pointer the runtime had allocated. Always append the host -fsanitize=address flags, and fail the build if the installed amdhip64 is missing __asan_init.
ccache drops -Xarch_* unless the compile also passes -arch, so TheRock's -Xarch_host -fsanitize=address never reaches clang and libamdhip64 stays uninstrumented. Attach a plain host -fsanitize=address to the CLR C/C++ compiles, including amdhip64 and rocclr.
dileepr1
added
the
ci:host-asan
dileepr1
marked this pull request as ready for review
October 21, 2026 15:34
dileepr1
added
ci:host-asan
The pull request was unmergeable, so pull_request CI never started. Keep the test workflow that pins the host-ASAN TheRock branch and runs hip-tests and rocrtst.
dileepr1
marked this pull request as draft
October 21, 2026 15:36
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
| jobs: | ||
| setup: | ||
| uses: ROCm/TheRock/.github/workflows/setup_multi_arch.yml@47434387de730d51788b705e5600602dd6064997 # 2026-10-16 | ||
| uses: ROCm/TheRock/.github/workflows/setup_multi_arch.yml@users/dravindr/host_asan_8455_test # 2026-09-30 |
| needs.setup.outputs.enable_build_jobs == 'true' | ||
| }} | ||
| uses: ROCm/TheRock/.github/workflows/multi_arch_ci_linux.yml@47434387de730d51788b705e5600602dd6064997 # 2026-10-16 | ||
| uses: ROCm/TheRock/.github/workflows/multi_arch_ci_linux.yml@users/dravindr/host_asan_8455_test # 2026-09-30 |
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Validation branch, not for merge.
Re-runs the runtime-only host-ASAN presubmit against ROCm/TheRock#8455 instead of ROCm/TheRock#7780, to check that #8455 reaches the same result by a different route.
#7780 added an
asan_presubmitworkflow input and enabled presubmit tests only whenexternal_repo and asan_presubmitwere both set. #8455 has no such condition:host-asanlistspull_requestin itstest_triggersinamdgpu_family_matrix.py, so any host-ASAN presubmit build runs its tests. The input is gone from this workflow as a result; nothing else about the run changed.TheRock side is
users/dravindr/host_asan_8455_test, cut fresh frommain(d7df0e20b) and carrying #8455, #8341 (ASAN runtime resolution), and the temporary commit that lets component tests run past the failing sanity check. The libraries-baseline-gate fix that the previous test branch carried is now merged inmain, so it is no longer applied on top.What to look for:
Linux::host-asanbuildscompiler-runtimeandruntime-testsongfx94X, andhip-testsplusrocrtstactually execute on the ASAN sandbox runner rather than being skipped.Supersedes #10088.