Skip to content

Revert "Remove comgr delay-load stub; statically link LLVM into comgr"#3877

Closed
chiranjeevipattigidi wants to merge 1 commit into
mainfrom
revert-3731-users/stellaraccident/remove-comgr-delay-load
Closed

Revert "Remove comgr delay-load stub; statically link LLVM into comgr"#3877
chiranjeevipattigidi wants to merge 1 commit into
mainfrom
revert-3731-users/stellaraccident/remove-comgr-delay-load

Conversation

@chiranjeevipattigidi
Copy link
Copy Markdown
Contributor

@chiranjeevipattigidi chiranjeevipattigidi commented Mar 10, 2026

Reverts #3731
Jax (v0.8.2 , v0.9.0) builds are failing due missing of libamd_comgr_stub.a during linking process
more details: #3876

Test Plan

Testing: https://github.com/ROCm/TheRock/actions/runs/22904894874

Test Result

TBD

@chiranjeevipattigidi
Copy link
Copy Markdown
Contributor Author

chiranjeevipattigidi commented Mar 10, 2026

there is another PR #3872 to fix the small bug(missed removing trailing comma in a json file) introduced by original pr.
If we agree to merge this revert , can close the #3872.

@chiranjeevipattigidi chiranjeevipattigidi marked this pull request as ready for review March 10, 2026 13:39
Copy link
Copy Markdown
Member

@marbre marbre left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ERROR: /root/.cache/bazel/_bazel_root/35c03de1b96cb539bbf04926e943e4d4/external/jax/jaxlib/rocm/BUILD:387:19: Linking external/jax/jaxlib/rocm/_hybrid.so failed: missing input file '@@local_config_rocm//rocm:rocm_dist/lib/libamd_comgr_stub.a'

seems like this rather needs a fix in JAX or at some other place. The revert is not the right long term fix. Before accepting a temporarily revert we should triage the root cause more carefully.

@stellaraccident stellaraccident self-requested a review March 10, 2026 13:42
@stellaraccident
Copy link
Copy Markdown
Collaborator

Neither of these are good candidates for revert. For build issues not covered by CI, please fix forward (and consider how to add CI coverage so it doesn't happen again).

Copy link
Copy Markdown
Collaborator

@stellaraccident stellaraccident left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

See above comment. Fix these other issues forward.

@stellaraccident
Copy link
Copy Markdown
Collaborator

ERROR: /root/.cache/bazel/_bazel_root/35c03de1b96cb539bbf04926e943e4d4/external/jax/jaxlib/rocm/BUILD:387:19: Linking external/jax/jaxlib/rocm/_hybrid.so failed: missing input file '@@local_config_rocm//rocm:rocm_dist/lib/libamd_comgr_stub.a'

seems like this rather needs a fix in JAX or at some other place. The revert is not the right long term fix. Before accepting a temporarily revert we should triage the root cause more carefully.

Jax has a really fragile manual list of libraries that needs update. Until that is fixed in a more robust way, we can't block landing changes like this (there is another one coming today that will also change the list of libraries Jax must bundle).

@github-project-automation github-project-automation Bot moved this from TODO to Done in TheRock Triage Mar 12, 2026
@chiranjeevipattigidi chiranjeevipattigidi deleted the revert-3731-users/stellaraccident/remove-comgr-delay-load branch March 31, 2026 04:02
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

Status: Done

Development

Successfully merging this pull request may close these issues.

3 participants