Enable gfx1250-strict in broad GPU families by marbre · Pull Request #8350 · ROCm/TheRock

marbre · 2026-09-18T16:12:29Z

Motivation

Family selection must include both gfx1250 targets for builds, distributions, and artifact fetching. Default test compilation includes strict along with the other registered targets. Explicit individual target selections retain their own payload identity.

Closes #8278

Technical Details

Prerequisites before landing:

~~Share bootstrap target ownership with SDK packaging #8325: shared SDK ownership.~~ (superseded)
Centralize target ownership metadata within TheRock #8377: centralized target ownership.
Group native kpack packages by target owner #8327: native package ownership.
Group target tarballs by package owner #8328: tarball ownership.
~~A rocm-systems submodule bump containing Share gfx1250 package ownership across bootstrap and kpack rocm-systems#11801.~~ (landed)
Established broader component readiness for family enablement.

Test Plan

Fresh compiler/HIP/RAND builds and SDK, DEB/RPM, and tarball validation passed on the integration branch with temporary family membership. The exact family enablement was then integrated for a combined build with RAND tests and target-neutral hipthreads. Test target overrides were left empty to exercise the default registered-target list. These checks do not establish full component or GPU execution support.

python -m pytest -q -p no:cacheprovider build_tools/tests/cmake_amdgpu_targets_test.py build_tools/tests/artifact_manager_tool_test.py build_tools/github_actions/tests/configure_multi_arch_ci_test.py
cmake --build ../therock-build-rand-tests-gfx125X-dcgpu --target artifact-rand artifact-hipthreads
../therock-build-rand-tests-gfx125X-dcgpu/compiler/amd-llvm/dist/lib/llvm/bin/llvm-objdump --offloading ../therock-build-rand-tests-gfx125X-dcgpu/math-libs/hipthreads/stage/lib/hipthreads/libhipthreads.a
python -m rocm_kpack.tools.verify_artifacts --artifacts-dir ../therock-build-rand-tests-gfx125X-dcgpu/artifacts --verbose --clang-offload-bundler ../therock-build-rand-tests-gfx125X-dcgpu/compiler/amd-llvm/dist/lib/llvm/bin/clang-offload-bundler

Test Result

171 passed and 1 existing test skipped with umask 022.
rocRAND and hipRAND tests compiled for both gfx1250 targets. Each target's test kpack contains 34 entries with only its target identity.
Target-neutral hipthreads compiled with the default test target list. Its library contains exactly 29 distinct device images, including separate gfx1250 and gfx1250-strict images.
All four artifact verifier checks passed.
No GPU tests were executed.

Submission Checklist

Look over the contributing guidelines at https://github.com/ROCm/TheRock/blob/main/CONTRIBUTING.md.

Generated with Codex

Family selection must include both gfx1250 targets for builds, distributions, and artifact fetching. Default test compilation includes strict along with the other registered targets. Explicit individual target selections retain their own payload identity. Prerequisites before landing: - #8325: shared SDK ownership. - #8327: native package ownership. - #8328: tarball ownership. - A rocm-systems submodule bump containing ROCm/rocm-systems#11801. - Established broader component readiness for family enablement. Fresh compiler/HIP/RAND builds and SDK, DEB/RPM, and tarball validation passed on the integration branch with temporary family membership. The exact family enablement was then integrated for a combined build with RAND tests and target-neutral hipthreads. Test target overrides were left empty to exercise the default registered-target list. These checks do not establish full component or GPU execution support. Closes #8278 Test plan: - python -m pytest -q -p no:cacheprovider build_tools/tests/cmake_amdgpu_targets_test.py build_tools/tests/artifact_manager_tool_test.py build_tools/github_actions/tests/configure_multi_arch_ci_test.py - cmake --build ../therock-build-rand-tests-gfx125X-dcgpu --target artifact-rand artifact-hipthreads - ../therock-build-rand-tests-gfx125X-dcgpu/compiler/amd-llvm/dist/lib/llvm/bin/llvm-objdump --offloading ../therock-build-rand-tests-gfx125X-dcgpu/math-libs/hipthreads/stage/lib/hipthreads/libhipthreads.a - python -m rocm_kpack.tools.verify_artifacts --artifacts-dir ../therock-build-rand-tests-gfx125X-dcgpu/artifacts --verbose --clang-offload-bundler ../therock-build-rand-tests-gfx125X-dcgpu/compiler/amd-llvm/dist/lib/llvm/bin/clang-offload-bundler Test results: - 171 passed and 1 existing test skipped with umask 022. - rocRAND and hipRAND tests compiled for both gfx1250 targets. Each target's test kpack contains 34 entries with only its target identity. - Target-neutral hipthreads compiled with the default test target list. Its library contains exactly 29 distinct device images, including separate gfx1250 and gfx1250-strict images. - All four artifact verifier checks passed. - No GPU tests were executed.

therock-pr-bot · 2026-09-18T16:13:05Z

✅ All Checks Passed — Ready for Review

Check	Status	Details
📝 PR Description	✅ Pass	—
⛔ Forbidden Files	✅ Pass	—
🧪 Unit Test	✅ Pass	PR does not contain code files — Unit Test auto-passed
🔎 pre-commit	✅ Pass	—
🚫 Draft PR	🔜 To Be Enabled	—
🚩 Feature Flag	🔜 To Be Enabled	—
📊 Code Coverage	🔜 To Be Enabled	—
🤖 therock-pr-bot	✅ Pass	—

🎉 All checks passed! This PR is ready for review.

📖 Need help? See the Policy FAQ for details on every check and how to fix failures.

🙋 Wish to Override Policy?

therock-pr-bot · 2026-09-18T16:13:38Z

🎉 All checks passed! This PR is ready for review.

ronlieb · 2026-09-19T10:54:35Z

        self.assertEqual(
-            self._select("gfx125X-all;gfx1250-strict"),
+            self._select("gfx125X-all;gfx942"),
            {


gfx942 added for gfx1250 patch ????

This looks fine. It's a unit test, and now that gfx1250-strict is part of gfx1250X-all, gfx942 is an "explicit family" supplementing the "broad family" (unit test name).

ronlieb · 2026-09-19T10:54:48Z

                "rand_lib_gfx125X-all.tar.zst",
                "rand_lib_gfx1250.tar.zst",
                "rand_lib_gfx1250-strict.tar.zst",
+                "rand_lib_gfx942.tar.zst",


same : gfx942 added for gfx1250 patch ????

Yes, this is tests for broader family enablement.

marbre · 2026-10-13T16:51:06Z

-- Configured AMDGPU Targets:
--   * gfx1250-strict : AMD Instinct gfx1250 strict
--   * gfx1250 : AMD Instinct MI450/MI450X/MI455X CDNA
--   Dist targets: gfx942;gfx1100;gfx1101;gfx1102;gfx1103;gfx1151;gfx1200;gfx1201;gfx1250-strict;gfx1250
--   Test targets: gfx900;gfx90c;gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1010;gfx1011;gfx1012;gfx1030;gfx1031;gfx1032;gfx1033;gfx1034;gfx1035;gfx1036;gfx1100;gfx1101;gfx1102;gfx1103;gfx1150;gfx1151;gfx1152;gfx1153;gfx1200;gfx1201;gfx1250-strict;gfx1250
--   * Dist bundle: gfx125X-dcgpu

https://github.com/ROCm/TheRock/actions/runs/38419442156/job/115384411886?pr=8350#logs

math-libs stage for gfx125X is passing.

JIRA ID : N/A ## Motivation - Run hipBLASLt + TensileLite on the gfx1250-strict path (#11777) for every PR. - Per-PR TheRock builds don't produce gfx1250-strict yet (ROCm/TheRock#8350, draft), so the job builds from source instead of waiting for artifacts. - Add Nightly schedule ## Technical Details - Toolchain: pinned ROCm SDK nightly (`rocm_sdk_version`, default `10.2.0a20261007`). A warn-only step flags when the SDK's compiler differs from the one at ci-env's `therock-ref`. - Checks that the runtime reports `gfx1250-strict` (`HSA_DISABLE_GFX12_STRICT=0`), then builds with `install.sh -a gfx1250-strict --skip_rocroller`. - One runner per GPU: the job container gets the runner's `ROCR_VISIBLE_DEVICES` (`-e ROCR_VISIBLE_DEVICES`) and fails early unless it names exactly one GPU. Checked on the host during a CI run: all 4 hipBLASLt shards ran on the runner's own GPU. - hipBLASLt: loads `library/gfx1250-strict/`; fails if `hipblaslt-test` doesn't link the library built here or its code objects aren't labelled `gfx1250-strict`. Runs as 4 gtest shards. - TensileLite: Uses 4 workers sharing the GPU. Unit tests dropped (host-only, already run per PR in TheRock CI). Selection `-m "common and not skip-gfx1250-strict"`; skipped if setup failed, still runs after a hipBLASLt failure. - Log scrub on build and test output: hardware details and wall-clock times are removed before anything reaches the console or the uploaded logs. - Per PR: `pull_request` trigger behind a GitHub-hosted gate (non-draft, ROCm org label, changes under `projects/hipblaslt`, its shared deps, this workflow or `ci-env/action.yml`). One run per PR; a newer push cancels it. Manual dispatch and the nightly are ungated. A manual run and the nightly on the same branch don't cancel each other. - Nightly: develop at 03:00 UTC (7 pm PST, 8 pm PDT). - Runner state: a fresh venv per job from the image's Python 3.12 (no `actions/setup-python`, so nothing from the runner's tool cache); the runner's pip cache is used only by the SDK install and is read-only except for the nightly; the job hands the root-owned workspace back to the runner's user at the end. - Includes #13464 (merged into this branch): `skip-gfx1250-strict` on 4 TensileLite configs and the bf8 MX gtest rows limited to gfx950, + fixed comment saying gfx1250 has no MX kernels with bf8 inputs. - Skips `gsu_gfx1250-strict.yaml` on gfx1250-strict for now (`skip-gfx1250-strict`): it intermittently returns wrong values from GSU multi-buffer kernels (see Test Result). Followed up with the config owner; the skip comes out when that's fixed. - Removed: the `workflow_run` trigger, artifact resolution, commit-status posting, unused `GFX1250_HW_XFAIL`. ## Test Plan - This PR's own `gfx1250-strict hipBLASLt + TensileLite` check on the gfx1250 runners (the PR changes the workflow, so it runs here). ## Test Result - CI on the gfx1250 runners ([run 38319193893](https://github.com/ROCm/rocm-libraries/actions/runs/38319193893)): hipBLASLt 68,138 tests over 4 shards, 0 failed; TensileLite common 109 passed, 2 xfailed, 0 failed. ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/TheRock/blob/main/GOVERNANCE.md#pull-requests. ## Risk level - Low. Self-contained workflow. PR code runs on a self-hosted runner only for ROCm org-labelled PRs 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude <noreply@anthropic.com>

## Motivation When multiple selected GPU targets share one native package owner, split artifact manifests can retain source roots whose device images have moved into a shared kpack under another root. The grouped-root validation introduced by #8327 rejects those absent directories before the DEB and RPM backends can apply their established filtering. The device-package builder then converts that exception into an empty result. Other variants can still be produced, causing package generation to appear successful and emit meta packages that depend on device packages that were never created. The failure is only discovered later during simulated or native package installation. This was first reported in #8327 (comment) and reproduces when #8350 enables `gfx1250-strict` alongside `gfx1250`. This PR carries forward the fix and artifact evidence from #8702, then adds RPM coverage and makes unexpected partial package failures fatal. Progress on #8278. ## Evidence The original #8702 investigation used https://github.com/ROCm/TheRock/actions/runs/36947069646. Both sparse device artifact manifests retained the rocSPARSE root after their device images moved into shared kpacks under hipSPARSELt. Each target kpack still contained 340 rocSPARSE and 3 hipSPARSELt images, while the generic artifact contained `librocsparse.so`. The current failure reproduces in #8350 run https://github.com/ROCm/TheRock/actions/runs/38419442156: - DEB package generation fails its simulated installation after the affected `gfx1250` device packages are omitted: https://github.com/ROCm/TheRock/actions/runs/38419442156/job/115420436430?pr=8350 - RPM generation completes despite omitting the affected packages: https://github.com/ROCm/TheRock/actions/runs/38419442156/job/115420436578?pr=8350 - The downstream RHEL 8 and SLES 16 installation jobs then fail on unresolved dependencies: - https://github.com/ROCm/TheRock/actions/runs/38419442156/job/115430597884?pr=8350 - https://github.com/ROCm/TheRock/actions/runs/38419442156/job/115430596637?pr=8350 The omitted packages include `amdrocm-sparse`, `amdrocm-blas-test`, `amdrocm-rand-test`, `amdrocm-ccl-test`, and `amdrocm-rccl-test` for the shared `gfx1250` package owner. ## Technical Details - Ignore absent manifest roots during kpack artifact discovery, matching the existing artifact-catalog behavior. - Continue collision validation for every existing package root. - Log skipped roots at debug level so stale manifest entries remain diagnosable. - Preserve the selected `gfx1250` and `gfx1250-strict` kpack payloads in their shared `gfx1250` native package. - Propagate unexpected device-package exceptions instead of converting them to empty results. - Keep intentional empty device variants skippable. - Return a nonzero status after the build summary whenever any base package is recorded as failed. - Add matching DEB and RPM regressions for combined-owner packages with absent manifest roots. The implementation is architecture-independent packaging logic. Regression fixtures use the motivating combined `gfx1250` and `gfx1250-strict` ownership case and verify that both selected payloads survive. No GPU execution is required for this change. ## Test Plan - `python build_tools/packaging/linux/tests/build_package_test.py` - `gfx125X,gfx1250-strict` dev build: https://github.com/ROCm/TheRock/actions/runs/38459803112 ## Test Result - 52 packaging tests passed. - The DEB regression verifies both selected kpack filenames and bytes survive staging when manifests contain absent roots. - The RPM regression verifies both valid target roots are retained in the generated spec and absent roots are excluded. - Device-package exceptions propagate, while intentional empty variants remain skippable. - Generic recorded package failures exit with status 1 after printing the build summary. ## Related - Tracking issue: #8278 - Original fix and evidence: #8702 - Native package ownership change: #8327 - Reproducing family enablement: #8350 ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/TheRock/blob/main/CONTRIBUTING.md. Generated with Codex --------- Co-authored-by: Ammar ELWazir <3832908+ammarwa@users.noreply.github.com>

ScottTodd · 2026-10-14T19:52:45Z

PyTorch build failures: https://github.com/ROCm/TheRock/actions/runs/38587475275/job/115983081260?pr=8350#step:15:38204

/__w/TheRock/TheRock/external-builds/pytorch/pytorch/aten/src/ATen/../../../third_party/composable_kernel/include/ck/utility/amd_buffer_addressing_builtins.hpp:48:48: error: use of undeclared identifier 'CK_BUFFER_RESOURCE_3RD_DWORD'
   48 |     wave_buffer_resource.config(Number<3>{}) = CK_BUFFER_RESOURCE_3RD_DWORD;
      |                                                ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
/__w/TheRock/TheRock/external-builds/pytorch/pytorch/aten/src/ATen/../../../third_party/composable_kernel/include/ck/utility/amd_buffer_addressing_builtins.hpp:61:22: error: use of undeclared identifier 'CK_BUFFER_RESOURCE_3RD_DWORD'
   61 |     auto flags     = CK_BUFFER_RESOURCE_3RD_DWORD;
      |                      ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
/__w/TheRock/TheRock/external-builds/pytorch/pytorch/aten/src/ATen/../../../third_party/composable_kernel/include/ck/utility/amd_buffer_addressing_builtins.hpp:73:22: error: use of undeclared identifier 'CK_BUFFER_RESOURCE_3RD_DWORD'
   73 |     auto flags     = CK_BUFFER_RESOURCE_3RD_DWORD;
      |                      ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
4 errors generated when compiling for gfx1250-strict.

Might need ROCm/pytorch#3639 or something similar on each pytorch release branch.

It exists only because TheRock multi-arch PR builds do not produce gfx1250-strict artifacts. Planned for removal once ROCm/TheRock#8350 lands and gfx1250-strict runners are available in TheRock CI. Co-Authored-By: Claude <noreply@anthropic.com>

…ite timeout; cleanup (#13771) JIRA ID : N/A ## Motivation Follow-ups from the #12983 and #13464 reviews, split out of #13620, plus a fix for hipBLASLt tests running ~7x slower since per-PR runs started. ## Technical Details - **hipBLASLt test speed:** `OMP_NUM_THREADS` and `OPENBLAS_NUM_THREADS` = host CPUs / (4 runners x 4 shards); `BLIS_NUM_THREADS` = the same, at most 8 (hipBLASLt's own BLIS default, which setting `OMP_NUM_THREADS` turns off). By default the CPU reference's OpenMP and OpenBLAS pools each start one thread per host CPU, so 4 runners x 4 shards oversubscribe the host 16x per pool: since per-PR runs started (10-12 to 10-15), the step took ~80 min when other runners were testing vs ~15 min alone, up to 110 min of its 120-min timeout. - **Gate:** add `cmake/` (rocisa and stinkytofu include `cmake/modules/default_amdclang.cmake`). - **One-GPU check:** runs first in its step, under `set -euo pipefail`. `ROCR_VISIBLE_DEVICES` comes from the runner's `.env`; if it is unset, N comes from the runner name's `-gpuN` (the pool registers runner N as `...-gpuN` with `ROCR_VISIBLE_DEVICES=N`), with a warning. It must be a single GPU index (`^[0-9]+$`), or the job fails before the build. Drop `HIP_VISIBLE_DEVICES` (never forwarded into the container). - SDK pinned to `10.2.0a20261013` (compiler `c7ceccf9a5`). Per-PR CI has since moved to compiler `bdc35841c9` (#13835, 10-19), so the workflow's compiler check now warns; re-pinning to `10.2.0a20261018` or later, which carries `bdc35841c9`, is a follow-up. - **upload-artifact:** v4.6.2 -> v7.0.2 (Node 20 deprecated). v7 needs Actions Runner >= 2.327.1; the gfx1250 runner is 2.337.0. - **TensileLite per-test timeout:** 600 -> 1200 s, this workflow only (FFM and TheRock keep their own). `decouple_pgr_tdm_fuse_gfx1250.yaml` has timed out at 600 s in most runs since #12825 grew its sweep 4.5x; before, it passed 41 of 41. - **Header:** the workflow is marked temporary, to be removed once TheRock's multi-arch builds produce gfx1250-strict artifacts (ROCm/TheRock#8350) and gfx1250-strict runners are available in TheRock CI. ## Test Plan - OpenMP cap: 4 jobs x 4 shards of a `*pre_checkin*` slice on a 128-core gfx942 host, default vs capped. - One-GPU check: 7 values under `bash -euo pipefail`. ## Test Result - [x] Run https://github.com/ROCm/rocm-libraries/actions/runs/38994215757: all steps green with no warnings on SDK 10.2.0a20261013 - hipBLASLt 68,925 tests, 0 failed, in 3.9 min (OpenMP/OpenBLAS capped, BLIS at most 8) while an uncapped run on the same host took 52 min - TensileLite 113 passed, including decouple_pgr_tdm_fuse_gfx1250.yaml under the 1200 s timeout. ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/TheRock/blob/main/GOVERNANCE.md#pull-requests. ## Risk level Low: CI-only, gfx1250 workflow. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude <noreply@anthropic.com>

scchan · 2026-10-22T22:42:33Z

We haven't enabled the reporting of gfx1250-strict by default yet (ROCm/rocm-systems#11575). You'll have to set env var HSA_DISABLE_GFX12_STRICT=1 to enable it. I presume you'd have to set that in the environment running tests for gfx1250-strict?

marbre · 2026-10-23T04:02:52Z

We haven't enabled the reporting of gfx1250-strict by default yet (ROCm/rocm-systems#11575). You'll have to set env var HSA_DISABLE_GFX12_STRICT=1 to enable it. I presume you'd have to set that in the environment running tests for gfx1250-strict?

I don't think this is related to the work in this PR. This is the family enablement in builds and it does not enable running tests.

scchan · 2026-10-23T14:27:59Z

We haven't enabled the reporting of gfx1250-strict by default yet (ROCm/rocm-systems#11575). You'll have to set env var HSA_DISABLE_GFX12_STRICT=1 to enable it. I presume you'd have to set that in the environment running tests for gfx1250-strict?

I don't think this is related to the work in this PR. This is the family enablement in builds and it does not enable running tests.

The reason I asked is because your PR description contains test results that imply that some gfx1250-strict tests have been run.

marbre · 2026-10-23T23:28:48Z

The reason I asked is because your PR description contains test results that imply that some gfx1250-strict tests have been run.

Please check the PR description more carefully:

No GPU tests were executed.

This is pure build tests.

ScottTodd

CI results are looking good now that #9549 is merged to skip pytorch builds on gfx1250-strict.

github-project-automation Bot added this to TheRock Triage Sep 18, 2026

github-project-automation Bot moved this to TODO in TheRock Triage Sep 18, 2026

marbre requested a review from ScottTodd September 18, 2026 16:12

ronlieb reviewed Sep 19, 2026

View reviewed changes

ScottTodd mentioned this pull request Sep 25, 2026

Group native kpack packages by target owner #8327

Merged

1 task done

ammarwa pushed a commit that referenced this pull request Sep 26, 2026

Integrate #8350: Enable gfx1250-strict in broad GPU families

2545d8f

ammarwa pushed a commit that referenced this pull request Sep 26, 2026

Integrate #8350: Enable gfx1250-strict in broad GPU families

f78c4fd

ammarwa pushed a commit that referenced this pull request Sep 26, 2026

Integrate #8350: Enable gfx1250-strict in broad GPU families

d72f7b1

marbre closed this Sep 29, 2026

marbre reopened this Sep 29, 2026

ScottTodd mentioned this pull request Oct 9, 2026

Enable gfx1170/gfx1171 targets #8840

Merged

1 task

archana-ramalingam mentioned this pull request Oct 10, 2026

ci(gfx1250): self-build gfx1250-strict per PR ROCm/rocm-libraries#12983

Merged

1 task done

Merge branch 'main' into users/marbre/gfx1250-strict-family-enablement

f0d94bd

marbre marked this pull request as ready for review October 13, 2026 16:51

ScottTodd approved these changes Oct 13, 2026

View reviewed changes

marbre mentioned this pull request Oct 13, 2026

Fix native packaging for absent split-artifact roots #8987

Merged

1 task done

Merge branch 'main' into users/marbre/gfx1250-strict-family-enablement

6cec237

ScottTodd and others added 3 commits October 16, 2026 08:04

Merge branch 'main' into users/marbre/gfx1250-strict-family-enablement

492e6ef

Merge branch 'main' into users/marbre/gfx1250-strict-family-enablement

2b64452

Merge branch 'main' into users/marbre/gfx1250-strict-family-enablement

103c751

marbre mentioned this pull request Oct 20, 2026

ci(rocroller): skip rocroller-tests on gfx1250 #9377

Open

2 of 3 tasks

archana-ramalingam mentioned this pull request Oct 21, 2026

ci(gfx1250): cap OpenMP threads in hipBLASLt tests; increase tensilelite timeout; cleanup ROCm/rocm-libraries#13771

Merged

2 tasks done

Merge branch 'main' into users/marbre/gfx1250-strict-family-enablement

2a17e79

ScottTodd mentioned this pull request Oct 28, 2026

[bkc] Pin to use release/bkc/therock-10.3-20261027 ROCm/rocm-libraries#14471

Merged

1 task done

Merge branch 'main' into users/marbre/gfx1250-strict-family-enablement

e4df0dd

ScottTodd approved these changes Oct 30, 2026

View reviewed changes

marbre merged commit 682b340 into main Oct 30, 2026
206 of 210 checks passed

marbre deleted the users/marbre/gfx1250-strict-family-enablement branch October 30, 2026 05:09

geotseng-amd mentioned this pull request Nov 5, 2026

revert(hipblaslt): remove the gfx1250v0 subtree and the asicRevision check ROCm/rocm-libraries#15123

Open

1 task

Conversation

marbre commented Sep 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Motivation

Technical Details

Test Plan

Test Result

Submission Checklist

Uh oh!

therock-pr-bot Bot commented Sep 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✅ All Checks Passed — Ready for Review

Uh oh!

therock-pr-bot Bot commented Sep 18, 2026

Uh oh!

ronlieb Sep 19, 2026

Choose a reason for hiding this comment

Uh oh!

ScottTodd Sep 21, 2026

Choose a reason for hiding this comment

Uh oh!

marbre Sep 30, 2026

Choose a reason for hiding this comment

Uh oh!

ronlieb Sep 19, 2026

Choose a reason for hiding this comment

Uh oh!

marbre Sep 30, 2026

Choose a reason for hiding this comment

Uh oh!

marbre commented Oct 13, 2026

Uh oh!

ScottTodd commented Oct 14, 2026

Uh oh!

scchan commented Oct 22, 2026

Uh oh!

marbre commented Oct 23, 2026

Uh oh!

scchan commented Oct 23, 2026

Uh oh!

marbre commented Oct 23, 2026

Uh oh!

ScottTodd left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

marbre commented Sep 18, 2026 •

edited

Loading

therock-pr-bot Bot commented Sep 18, 2026 •

edited

Loading