Skip to content

Bump rocm-systems from 093b66c to 96e30b4#3984

Open
dependabot[bot] wants to merge 5 commits intomainfrom
dependabot/submodules/main/rocm-systems-96e30b4
Open

Bump rocm-systems from 093b66c to 96e30b4#3984
dependabot[bot] wants to merge 5 commits intomainfrom
dependabot/submodules/main/rocm-systems-96e30b4

Conversation

@dependabot
Copy link
Contributor

@dependabot dependabot bot commented on behalf of github Mar 16, 2026

Bumps rocm-systems from 093b66c to 96e30b4.

Commits
  • 96e30b4 call HIP_SKIP_TEST properly if device has no image support (#4040)
  • e1c229e SWDEV-581075 - Skip if managed memory not supported (#3598)
  • ce28e5f copy amd_icd_file to stage/etc/OpenCL/vendors (#3818)
  • 55b57ee [ci] disabling mi325 test runners (#4062)
  • c873a9b [rocprofiler-compute][tui] Align TUI's analysis logic to GUI (#3641)
  • 485f87f [rocprofiler-systems] Enable the roctx-api-runtime-instrument tests (#3990)
  • a8b23ed [TheRock CI] Fix workflow_dispatch project selection being overridden by CI f...
  • ceda961 [HSA][CLR][rocprofiler-sdk] hsa_amd_memory_async_batch_copy cherry-pick #3824...
  • d7d61f2 [rocdecode]users/rrawther/media libs test fix (#4056)
  • 45365fb [rocprofiler-systems] Fix the ubuntu-jammy workflow (#4055)
  • Additional commits viewable in compare view

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.


Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

  • @dependabot rebase will rebase this PR
  • @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
  • @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
  • @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
  • @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
  • @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)

Bumps [rocm-systems](https://github.com/ROCm/rocm-systems) from `093b66c` to `96e30b4`.
- [Release notes](https://github.com/ROCm/rocm-systems/releases)
- [Commits](ROCm/rocm-systems@093b66c...96e30b4)

---
updated-dependencies:
- dependency-name: rocm-systems
  dependency-version: 96e30b429c074f8a3fcf66be9c014df4d7cbc681
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>
@dependabot dependabot bot added dependencies Pull requests that update a dependency file submodules Pull requests that update submodules code labels Mar 16, 2026
@dependabot dependabot bot requested a review from a team as a code owner March 16, 2026 10:49
@dependabot dependabot bot added dependencies Pull requests that update a dependency file submodules Pull requests that update submodules code labels Mar 16, 2026
@chiranjeevipattigidi
Copy link
Contributor

It looks like the builds are still failing to apply the patch:
0003-Use-is_versioned-true-consistently-in-both-Comgr-Loa.patch
(Ref: https://github.com/ROCm/TheRock/blob/main/patches/amd-mainline/rocm-systems/0003-Use-is_versioned-true-consistently-in-both-Comgr-Loa.patch)
As mentioned earlier, @AmosLewis already shared some possible solutions in the previous PR discussion:
#3956 (comment)

@ScottTodd
Copy link
Member

Docs for how to work with patch files: https://github.com/ROCm/TheRock/tree/main/patches#resolving-conflicts-with-patches

@ScottTodd
Copy link
Member

The fix is to delete the patch. See ROCm/rocm-systems@b7167d0

Copy link
Contributor

@geomin12 geomin12 left a comment

Choose a reason for hiding this comment

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

patches need to be fixed!

rocm-systems at 96e30b4 includes commit b7167d0 which sets
kComgrVersioned = amd::IS_HIP (versioned for HIP, unversioned for OpenCL).
Per patches/README.md: when the conflicting commit is equivalent to the
patch, delete the patch. No longer apply this patch in TheRock CI.

Made-with: Cursor
@AmosLewis
Copy link
Contributor

AmosLewis commented Mar 17, 2026

Component Classification Error Job Link Filed Issues
hipsparse (Win gfx1151, 1/1 shard) Code Failure pre_checkin/bsrsv2.level2, csrsv2.level2 unit tests failed (f64_c, f32_c; block_dim 2/3/9/27/81) gfx1151 Possible overlap rocm-libraries #5502
rocm[libraries,devel] wheels (Linux gfx1151) Test/Packaging Bug ROCmDevelTest.testSharedLibrariesLoad: ctypes.CDLL passed directory etc/ld.so.conf.d → OSError "Is a directory" gfx1151 #3999
Windows gfx110X-all Build release Infra Self-hosted runner lost communication with server (~2h 20m) job

Recommend: failures are pre-existing test/infra issues, not introduced by rocm-systems bump 093b66c→96e30b4 or patch delete. hipsparse pattern may overlap with rocm-libraries #5502. rocm-sdk test should skip directories in testSharedLibrariesLoad. Re-run only for infra (gfx110X) if needed; merge acceptable with known test failures tracked elsewhere. cc @geomin12

@geomin12
Copy link
Contributor

Re-based as we now have gfx94X runners to test on

@dependabot @github
Copy link
Contributor Author

dependabot bot commented on behalf of github Mar 18, 2026

A newer version of rocm-systems exists, but since this PR has been edited by someone other than Dependabot I haven't updated it. You'll get a PR for the updated version as normal once this PR is merged.

@chiranjeevipattigidi
Copy link
Contributor

chiranjeevipattigidi commented Mar 18, 2026

For rccl test failures, existing issue: #3774

@amd-hsivasun
Copy link
Contributor

Reran tests as there appears to be some flaky tests. Namely rocfft, hipsparselt and rocgdb.

Copy link

@amd-aakash amd-aakash left a comment

Choose a reason for hiding this comment

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

Awaiting summary of known issues - lets merge this

@ScottTodd
Copy link
Member

#3999 should really be fixed before we merge this. I don't see anyone actually working on it yet though. It should be a 1 line change to a python test file...

@amd-hsivasun
Copy link
Contributor

#3999 should really be fixed before we merge this. I don't see anyone actually working on it yet though. It should be a 1 line change to a python test file...

The fix for this issue seems to have landed in 4 hours ago. We are starting another Bump PR cycle right after we merge this one.

@amd-hsivasun
Copy link
Contributor

amd-hsivasun commented Mar 20, 2026

Summary for known issues

Issues with workarounds (while a full fix is pending)

Component Issue Notes
rocm-sdk test #3999 PR #4015, Introduced some changes which need to be viewed in CI of this bump. See this comment #4015 (comment)
hipsparse #5502 Repro was delayed while waiting for a system. In the meantime, workaround is expected to merge via #5605.
libhipcxx TheRock#4031 Failure is not attributed to this bump; it showed up when MI300 machines were unavailable in an earlier nightly, likely tied to a compiler bump. Workaround direction: TheRock#4087.
rocprofiler-compute TheRock#4034 Test may be flaky; guidance from dev is not to run this test.

Issues with no update

Component Issue Notes
RCCL TheRock#3774 A fix was proposed in rocm-systems#3945 but it has remained unmerged for some time.

@ScottTodd
Copy link
Member

#3999 should really be fixed before we merge this. I don't see anyone actually working on it yet though. It should be a 1 line change to a python test file...

The fix for this issue seems to have landed in 4 hours ago. We are starting another Bump PR cycle right after we merge this one.

No? As pointed out on that PR, it does not fix the issue. It is related, but there is still a bug in the python tests.

@amd-hsivasun
Copy link
Contributor

#3999 should really be fixed before we merge this. I don't see anyone actually working on it yet though. It should be a 1 line change to a python test file...

The fix for this issue seems to have landed in 4 hours ago. We are starting another Bump PR cycle right after we merge this one.

No? As pointed out on that PR, it does not fix the issue. It is related, but there is still a bug in the python tests.

Sorry read through the PR but also saw "should also fix issue - #3999" in the PR description. Will hold off on Merging

Copy link

@amd-aakash amd-aakash left a comment

Choose a reason for hiding this comment

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

Lets address the changes mentioned by @ScottTodd - create a fresh Bump and merge that in after watching if we fix these issues.

@ScottTodd
Copy link
Member

I suspect #4094 will fix #3999. We're testing now. I'd say we don't necessarily need to sync/rebase/restart this submodule update PR after that, we can just merge the PRs in sequence. It's a test at the leaf of the packages (but would affect users who pip install then run rocm-sdk test - that's why it matters to not let it break and rely on fixing afterwards).

Copy link
Member

@ScottTodd ScottTodd left a comment

Choose a reason for hiding this comment

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

Approving the patch deletion for CODEOWNERS

The test failures in rocm-sdk test should be fixed now that #3999 is fixed (I tested with artifacts from this PR itself, see

I haven't personally been tracking the other issues spotted on this submodule update, so trusting others there.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

dependencies Pull requests that update a dependency file submodules Pull requests that update submodules code

Projects

None yet

Development

Successfully merging this pull request may close these issues.

6 participants