Fork configuration scripts and workflows for multi-arch CI by ScottTodd · Pull Request #4123 · ROCm/TheRock

ScottTodd · 2026-03-23T23:37:04Z

This adds new setup_multi_arch.yml and configure_multi_arch_ci_summary.py configuration code for multi-arch CI.

Motivation

The existing build_tools/github_actions/configure_ci.py script and .github/workflows/setup.yml workflow were both tightly coupled to the non-multi-arch CI pipelines and multi-arch CI has some unique needs:

[Multi-arch] Add stage-aware "use_prebuilt_artifacts" configuration to CI workflows #3399
Different workflow I/O (single-arch CI has a matrix across each family, multi-arch CI has a single pipeline per platform)
(Related) Setting alternate schedules, see [CI] Weekly CI and introducing a new amdgpu matrix generator #1732

I also judged that starting fresh would be easier, given architectural issues with the existing code:

Mixed responsibilities in a 300 LOC matrix_generator() function
Sequencing of decisions (whether to skip, what to build, what to test, etc.) was scattered and sometimes duplicated

Technical Details

Note

See my worklog for this feature branch here: tasks/active/multi-arch-configure.md

The new code is structured as a sequence of stages:

Parse inputs from github (triggers, labels) and git (files changed)
Check "skip CI" gate to early return
Decide which jobs to run and with what options (skip/use prebuilt/build)
Decide which GPU targets/families to build (trigger type + labels --> per-platform GPU families)
Expand per-platform build configuration data structures
Write outputs to github outputs and step summary

It includes several ✨NEW✨ features:

Rendered markdown summarizing the configure output

Before After

https://github.com/ROCm/TheRock/actions/runs/23509071878?pr=4142 https://github.com/ROCm/TheRock/actions/runs/23465566745?pr=4123
A collapsed build graph instead of a matrix with a single build variant

Before After

https://github.com/ROCm/TheRock/actions/runs/23509071878?pr=4142 https://github.com/ROCm/TheRock/actions/runs/23465566745?pr=4123
Reading inputs via GITHUB_EVENT_PATH instead of explicit/verbose github.event.inputs plumbing
Passing outputs via build_config JSON instead of explicit/verbose artifact_group, matrix_per_family_json, dist_amdgpu_families, etc. (to stay under input count limits and make cross-file maintenance hopefully easier)

As well as a few 🪦REMOVED 🪦 features:

No longer using determine_long_lived_branch() to change push behavior based on the branch name - if we enable the workflow on a branch we should run the same set of jobs
No run_functional_tests plumbing - this has not been added to multi-arch CI [yet?]

Comparison Metrics

📊 Feature Comparison

Feature	Old	New	Change
Pipeline architecture	Monolithic `matrix_generator` (295 lines)	6-step pipeline of pure functions	Redesigned
Typing	Untyped `base_args` dict	11 frozen dataclasses, `JobAction` enum	New
Skip CI gate	Buried in `matrix_generator` + `main()`	`should_skip_ci()`	Simplified
Target selection	Interleaved with matrix expansion	`select_targets()`	Separated
Test type	Post-hoc mutation in `main()`	`_determine_test_type()`	Separated
Matrix output	Array of per-variant rows via `strategy.matrix`	Single `build_config` JSON per platform	Simplified
Prebuilt stages	Boolean `use_prebuilt_artifacts`	Per-stage `dict[str, JobAction]` on `BuildRocmDecision`	More granular
Workflow contract	Implicit (YAML ↔ Python drift possible)	Contract tests extract `fromJSON` refs from YAML, assert against dataclass fields	Validated
Step summary	Limited information	Markdown with skip reasons, per-family test table, non-default callouts	Redesigned
`setup.yml` coupling	8+ env vars piped through YAML → Python	Script reads `GITHUB_EVENT_PATH` directly	Decoupled

🔎 Code Complexity

Around the same number of statements but with more comments and structure:

Metric	Old `configure_ci.py` (main)	New `configure_multi_arch_ci.py` + summary	Remaining `configure_ci.py`
Lines	828	947 + 204 = 1,151	676
Statements	348	335 + 114 = 449	302
Functions	8	30 + 6 = 36	7
Classes/dataclasses	0	11	0

🧪 Test Coverage

Logic and dataclasses are tested step by step, inputs/outputs are pushed to the boundaries for easier testing.

Metric	Old (main)	New	Delta
Test functions	33 (configure_ci_test)	47 (multi_arch)	+14
Test lines	907	826	-81
Remaining configure_ci tests	—	27	-6 (multi-arch tests removed)
Statement coverage (main script)	~63% (configure_ci.py)	84% (configure_multi_arch_ci.py)	+21pp
Uncovered code	`main()`, multi-arch paths mixed with single-arch	I/O boundary only (`from_environ`, `write_outputs`, `main`)	cleaner boundary

Test Plan

New unit tests
- Tests for each stage, using the dataclasses/enums/etc. that are passed between them
- A few integration tests for the whole configure() pipeline
- Tests for the workflow YAML files and their fromJSON usage
Manual testing
- on pull_request, multi-arch CI should run by default
- on push it should run with the same behavior as before
Watch multi-arch CI behavior for push events after merge

Test Result

Expected jobs ran on https://github.com/ROCm/TheRock/actions/runs/23773060560?pr=4123, configure output markdown looks as expected.

Submission Checklist

Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

New script for multi-arch CI configuration, designed as a pipeline of data transformations with typed dataclass interfaces between steps: 1. Parse Inputs (CIInputs from GITHUB_EVENT_PATH) 2. Check Skip CI (gate: skip-ci label, docs-only changes) 3. Select Targets (trigger type + labels → GPU families) 4. Decide Stages (changed files → rebuild/prebuilt per stage) 5. Expand Matrix (families × variant → matrix entries) 6. Write Outputs (JSON → GITHUB_OUTPUT) This commit is the scaffold — dataclasses and pipeline shape with stub implementations. Each step function has a TODO for Phase 2 logic. Tests (19 passing) demonstrate the pattern for exercising each step. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>

The CI pipeline is a DAG of job groups (build-rocm, test-rocm, build-rocm-python, build-pytorch, test-pytorch), not a flat list of stages. This restructures the data model to match: - JobGroupDecision base class (run/prebuilt/skip) with subclasses for groups that need extra details (BuildRocmDecision for per-stage granularity, TestRocmDecision for test type) - JobDecisions with explicit named fields for each job group node - decide_jobs() replaces decide_stages() - CIOutputs carries JobDecisions instead of flat test_type/stage lists 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>

- Rename internal field to is_ci_enabled (output key stays enable_build_jobs for workflow compat) - Remove from __future__ import annotations per style guide - Move inline imports to top of file - Use named arguments for multi-param function calls - Drop unused targets param from decide_jobs() - Note that steps 3 and 4 are independent 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>

- Collapse 4 event type property tests into 1 mutual exclusivity check - Extract _run_from_environ helper for GITHUB_EVENT_PATH test fixtures with doc link explaining the testing strategy - Remove _make_inputs helper — inline CIInputs construction so each test's inputs are visible at the call site - Use named args on expand_matrix test call - Focus from_environ tests on what's unique per event type (PR labels, workflow_dispatch inputs, push before SHA) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>

Reorder pipeline steps: decide_jobs (3) before select_targets (4) so target selection and matrix expansion are adjacent. select_targets policy per trigger type: - workflow_dispatch: explicit per-platform family lists from inputs - pull_request: presubmit defaults + gfx*/run-all-archs-ci label opt-ins - push: presubmit+postsubmit families - schedule: all families - Unknown trigger types raise ValueError Input parsing moved to the boundary: CIInputs.linux_amdgpu_families is now list[str] (parsed from comma-separated input in from_environ). Unknown family names fail fast with ValueError. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>

Expand comments in select_targets to explain the design rationale: trigger types form a progression from smallest (pull_request) to broadest (schedule), with workflow_dispatch giving full manual control. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

…matrix Port the multi-arch matrix expansion logic from configure_ci.py's generate_multi_arch_matrix into the new pipeline's expand_matrix function. Groups families by build variant into a single MatrixEntry with per-family JSON for downstream per-architecture job expansion. A parity test confirms identical output to the old function for the same inputs. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Changes: - Replace variant-grouping loop with a flat membership check — there's only ever one build_variant per workflow run - Add expand_matrices() returning a MatrixExpansion dataclass for both platforms, with _expand_matrix_for_platform as a private helper - Rename lookup_matrix → all_families for consistency with select_targets - Log when a platform has no config for the requested build variant Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Add amdgpu_family_matrix_test.py with schema validation tests: no duplicate family names per platform, required fields present, build_variants non-empty. This lets _expand_matrix_for_platform drop its defensive checks (seen_families dedup, .get() defaults for required fields) and index directly, relying on upstream validation from select_targets and data invariant tests. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Rename MatrixEntry → BuildConfig, expand_matrices → expand_build_configs. The helper returns BuildConfig | None instead of a 0-or-1-element list. BuildConfigs holds linux/windows as Optional fields. write_outputs now emits linux_build_config / windows_build_config as JSON objects (or empty string when skipped) plus linux_build_enabled / windows_build_enabled booleans, replacing the JSON array outputs that existed only to support a matrix strategy we don't need yet. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Test through the public expand_build_configs API, not the private per-platform helper. Verify structural properties (schema, consistency between dist_amdgpu_families and matrix_per_family_json, filtering behavior) rather than specific data values from amdgpu_family_matrix.py. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Check skip-ci PR label first (takes priority), then delegate to is_ci_run_required() from configure_ci_path_filters.py for docs-only / no-files-changed detection. schedule and workflow_dispatch always proceed (changed_files is None). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Add test_type determination to decide_jobs: quick (default), comprehensive (schedule), full (submodule changes or test labels), with test_filter: PR label override. Adopts the test type names from PR ROCm#3992. Extract GitContext dataclass to separate git-derived data from CIInputs. configure() is now fully pure — takes CIInputs + GitContext, no git calls. main() is the only place that builds GitContext.from_repo(). Tests construct GitContext directly. Rename parameters to ci_inputs/git_context throughout for clarity. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Replace mutable test_type/reason with early returns in priority order: test_filter label > test:* labels > schedule > submodule change > default. Eliminates risk of fall-through overwrites between conditions. Raise ValueError on unrecognized test_filter values instead of silently ignoring. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Each pipeline dataclass (CIInputs, GitContext, TargetSelection, JobDecisions, BuildConfigs) now has a log() method for CI diagnostics. configure() calls them after each step, replacing ad-hoc prints in main() and step functions. Remove getattr loops in favor of explicit field access. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Add setup_multi_arch.yml: checkout, call configure_multi_arch_ci.py (reads GITHUB_EVENT_PATH directly, no env var pass-through chain), compute package version. Update multi_arch_ci.yml to call setup_multi_arch.yml instead of setup.yml. Replace matrix strategy with direct fromJSON property access on per-platform build config objects. Route prebuilt_stages and baseline_run_id through the job graph model (BuildRocmDecision.stage_decisions and .baseline_run_id) rather than as raw pass-through strings. write_outputs derives the output values from the structured decisions. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

- Fix stale test_type values in module docstring (smoke/full → quick/ standard/comprehensive/full) - Replace sys.exit(1) with raise RuntimeError in from_environ() - Rename lookup_matrix → all_families in _filter_families_by_platform for consistency with the rest of the file - Remove unused sys import Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Lowercase family names in _parse_comma_list and PR labels in from_environ(). Matrix keys are lowercase (gfx94x) but workflow_dispatch descriptions show uppercase (gfx94X) — normalizing at the input boundary so internal functions don't need to care. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Replace the minimal bullet list with structured markdown: test type with reason, platform build config table (families, variant, artifact group), job group decision table, and prebuilt stages when set. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Rich step summary with DAG visualization, per-family test runner table, build-pytorch details, and non-default configuration callouts. Moved to a separate file to keep the main script focused on pipeline logic. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

…torch - Use GitHub alert syntax (> [!NOTE]) for non-default callouts - Mechanical trigger line: "Trigger: `event` on `branch`, `variant` variant." - "Runner Label" column header in test-rocm table - Add "# Build graph" comment to DAG - Remove build-pytorch section (not enough detail yet to be useful) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Add === phase headers === to configure() so logs read as a narrative. Print all CIInputs fields unconditionally (empty values are informative). Add context to check_skip_ci ("Checking N files against path filters"). Handle empty-families case in summary ("No GPU families selected"). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Replace 7+ individual BuildConfig inputs on multi_arch_ci_linux.yml and multi_arch_ci_windows.yml with a single build_config JSON string. Downstream workflows unpack with fromJSON(inputs.build_config).field. Rename matrix_per_family_json (double-encoded JSON string) to per_family_info (native list). Eliminates the fragile fromJSON(fromJSON(...)) pattern — matrix strategies now use fromJSON(inputs.build_config).per_family_info directly. Fix github_actions_utils → github_actions_api import after rebase. Remove stale parity test (matrix data diverged on main). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Verify that workflow YAML references to fromJSON(inputs.build_config) only use fields that exist on the Python BuildConfig dataclass. Linux checks exact match (all fields used), Windows checks no unknown fields (build_pytorch not yet used there). Also move imports to top of test file and update ci:skip/ci:run-all-archs label names. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Pass prebuilt stage info through expand_build_configs to _expand_build_config_for_platform, setting fields at construction time rather than mutating frozen dataclasses after the fact. Removes 2 separate outputs from setup_multi_arch.yml (9 → 7). Workflow YAML reads them from build_config via fromJSON. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

…ript Multi-arch CI now uses configure_multi_arch_ci.py exclusively. Remove generate_multi_arch_matrix, the multi_arch parameter from matrix_generator, and the ci:run-multi-arch label check from configure_ci.py. Add ci:run-multi-arch label gate to check_skip_ci in the new script: pull_request triggers require the label to opt in, avoiding doubled CI load during the transition. push/schedule/workflow_dispatch are unaffected. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

## Motivation I'm adding new code that depends on [`amdgpu_family_matrix.py`](https://github.com/ROCm/TheRock/blob/main/build_tools/github_actions/amdgpu_family_matrix.py) in #4123 and I want these structural invariants tested locally so I don't need to assert on them in the new downstream code. ## Technical Details We'll migrate over to https://github.com/ROCm/TheRock/blob/main/build_tools/github_actions/new_amdgpu_family_matrix.py soon, so I'm just adding a minimal set of tests for now. ## Test Plan * Run the unit tests ## Test Results Tests can fail as expected: * `test_required_fields_present` ```diff λ git diff diff --git a/build_tools/github_actions/amdgpu_family_matrix.py b/build_tools/github_actions/amdgpu_family_matrix.py --- a/build_tools/github_actions/amdgpu_family_matrix.py +++ b/build_tools/github_actions/amdgpu_family_matrix.py @@ -174,7 +174,6 @@ amdgpu_family_info_matrix_nightly = { "windows": { "test-runs-on": "", "family": "gfx900", - "fetch-gfx-targets": [], "build_variants": ["release"], "expect_pytorch_failure": True, }, ``` ``` (.venv) λ pytest build_tools/github_actions/tests/amdgpu_family_matrix_test.py -vv ... build_tools\github_actions\tests\amdgpu_family_matrix_test.py::TestFamilyMatrixInvariants::test_required_fields_present - AssertionError: gfx900/windows missing required fields: {'fetch-gfx-targets'} ``` * `test_no_duplicate_family_names_per_platform` ```diff diff --git a/build_tools/github_actions/amdgpu_family_matrix.py b/build_tools/github_actions/amdgpu_family_matrix.py index 63ad8c8..6eae4125 100644 --- a/build_tools/github_actions/amdgpu_family_matrix.py +++ b/build_tools/github_actions/amdgpu_family_matrix.py @@ -146,6 +146,18 @@ amdgpu_family_info_matrix_presubmit = { "build_variants": ["release"], }, }, + "gfx120x-all": { + "linux": { + # TODO(#2683): Re-enable label once stable + # Label is linux-gfx120X-gpu-rocm + "test-runs-on": "", + "family": "gfx120X-all", + "fetch-gfx-targets": ["gfx1200", "gfx1201"], + "bypass_tests_for_releases": True, + "build_variants": ["release"], + "sanity_check_only_for_family": True, + }, + }, } ``` ``` (.venv) λ pytest build_tools/github_actions/tests/amdgpu_family_matrix_test.py -vv ... build_tools\github_actions\tests\amdgpu_family_matrix_test.py::TestFamilyMatrixInvariants::test_no_duplicate_family_names_per_platform - AssertionError: Duplicate family 'gfx120X-all' on linux: target 'gfx120x-all' and 'gfx120x' ``` ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

HereThereBeDragons

looks good so far what i saw.

reviewed everything but:

build_tools/github_actions/configure_multi_arch_ci.py
build_tools/github_actions/tests/configure_multi_arch_ci_test.py

will do those tomorrow

HereThereBeDragons · 2026-03-26T20:26:07Z

+    return "\n".join(lines)
+
+
+def _non_default_callouts(ci_inputs: CIInputs, outputs: CIOutputs) -> list[str]:


i never heard callout in that context. maybe a few comments could help to set the expectation

Would "highlights" make more sense than "callouts"? These are any parts of the CI configuration that are different from what runs by "default", such as when a label was added to opt-in to some extra builds or tests. I figured that if we only showed tables with the computed values, it wouldn't be obvious what about those values is worth paying attention to.

ping pong with claude suggests:
_config_overrides
_override_notes
_active_overrides

## Motivation Progress on #3399. I'm extending the configuration scripts/workflows for multi-arch CI in #4123 and the docs could use a refresh. ## Technical Details * Highlight the differences between what jobs run on different triggers (I also have changes on #4123 that will surface this in the step summary output) * Elaborate on "CI" vs "Multi-Arch CI" setup and roadmap ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests. --------- Co-authored-by: Claude <noreply@anthropic.com>

ScottTodd · 2026-03-27T22:34:45Z

Thanks for the feedback. We've been landing substantial changes to the old/current configure_ci.py and workflows as part of switching our default workflow from ci.yml to multi_arch_ci.yml. I'll sync this, resolve merge conflicts, then start addressing the comments next week.

HereThereBeDragons

reviewing: build_tools/github_actions/configure_multi_arch_ci.py
i would recommend splitting this in multiple files. you already have 5 distinct steps + the data classes.

still missing: test file

see also comment about running those workflows for release branches.

Resolved conflicts and moved code for * log links * skip labels

Pass CIInputs to expand_build_configs instead of just build_variant, giving it access to PR labels. When a test_runner:<kernel> label is present (e.g. test_runner:oem), override the test runner for families that have a matching test-runs-on-kernel entry in the family matrix. Families without kernel support get their test runner cleared. Also adds test_runner: to the step summary highlights section. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>

ScottTodd · 2026-03-30T23:42:10Z

i would recommend splitting this in multiple files. you already have 5 distinct steps + the data classes.

I think splitting files will make more sense once we start filling in new logic for the build and test jobs. As is, this is just porting the existing configure_ci.py logic to a new framework.

ScottTodd · 2026-03-30T23:43:16Z

Addressed all comments. PTAL?

## Motivation I'm adding new code that depends on [`amdgpu_family_matrix.py`](https://github.com/ROCm/TheRock/blob/main/build_tools/github_actions/amdgpu_family_matrix.py) in #4123 and I want these structural invariants tested locally so I don't need to assert on them in the new downstream code. ## Technical Details We'll migrate over to https://github.com/ROCm/TheRock/blob/main/build_tools/github_actions/new_amdgpu_family_matrix.py soon, so I'm just adding a minimal set of tests for now. ## Test Plan * Run the unit tests ## Test Results Tests can fail as expected: * `test_required_fields_present` ```diff λ git diff diff --git a/build_tools/github_actions/amdgpu_family_matrix.py b/build_tools/github_actions/amdgpu_family_matrix.py --- a/build_tools/github_actions/amdgpu_family_matrix.py +++ b/build_tools/github_actions/amdgpu_family_matrix.py @@ -174,7 +174,6 @@ amdgpu_family_info_matrix_nightly = { "windows": { "test-runs-on": "", "family": "gfx900", - "fetch-gfx-targets": [], "build_variants": ["release"], "expect_pytorch_failure": True, }, ``` ``` (.venv) λ pytest build_tools/github_actions/tests/amdgpu_family_matrix_test.py -vv ... build_tools\github_actions\tests\amdgpu_family_matrix_test.py::TestFamilyMatrixInvariants::test_required_fields_present - AssertionError: gfx900/windows missing required fields: {'fetch-gfx-targets'} ``` * `test_no_duplicate_family_names_per_platform` ```diff diff --git a/build_tools/github_actions/amdgpu_family_matrix.py b/build_tools/github_actions/amdgpu_family_matrix.py index 63ad8c8..6eae4125 100644 --- a/build_tools/github_actions/amdgpu_family_matrix.py +++ b/build_tools/github_actions/amdgpu_family_matrix.py @@ -146,6 +146,18 @@ amdgpu_family_info_matrix_presubmit = { "build_variants": ["release"], }, }, + "gfx120x-all": { + "linux": { + # TODO(#2683): Re-enable label once stable + # Label is linux-gfx120X-gpu-rocm + "test-runs-on": "", + "family": "gfx120X-all", + "fetch-gfx-targets": ["gfx1200", "gfx1201"], + "bypass_tests_for_releases": True, + "build_variants": ["release"], + "sanity_check_only_for_family": True, + }, + }, } ``` ``` (.venv) λ pytest build_tools/github_actions/tests/amdgpu_family_matrix_test.py -vv ... build_tools\github_actions\tests\amdgpu_family_matrix_test.py::TestFamilyMatrixInvariants::test_no_duplicate_family_names_per_platform - AssertionError: Duplicate family 'gfx120X-all' on linux: target 'gfx120x-all' and 'gfx120x' ``` ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

HereThereBeDragons

thanks you. looks good.

only minor ideas for improvement, directly answered in the comments/your reply.

ScottTodd · 2026-03-31T17:36:01Z

Test Plan

Watch multi-arch CI behavior for push events after merge

Looks good: https://github.com/ROCm/TheRock/actions/runs/23810551251.

Same set of targets, test types, etc. as on the pull request,
Trigger detected as "push"

ScottTodd and others added 30 commits March 23, 2026 13:19

Sanitize some inputs with .lower().

be23fa5

Formatting cleanup

4d314b1

Fix summary output with default/empty workflow_dispatch inputs

027ed0c

Adapt to ci label changes

cabdc8e

HereThereBeDragons reviewed Mar 26, 2026

View reviewed changes

ScottTodd mentioned this pull request Mar 27, 2026

[ci] Multi-arch CI does not honor run-full-tests-only flag, while CI does #4184

Closed

HereThereBeDragons reviewed Mar 30, 2026

View reviewed changes

ScottTodd and others added 11 commits March 30, 2026 13:07

Merge remote-tracking branch 'upstream/main' into multi-arch-configure

19c7709

Resolved conflicts and moved code for * log links * skip labels

Fix tests after merging

efb6c33

Add build variant label to job names

d481320

Address comments in configure_multi_arch_ci_summary.py

5cc463e

Simplifying and documenting configure_multi_arch_ci.py fields

26d721a

Test fixes

f1a7488

Address more review comments in configure_multi_arch_ci.py

0856078

Merge remote-tracking branch 'upstream/main' into multi-arch-configure

6e38a1d

Add executable bit to configure scripts (??? why did it get dropped?)

84ab305

Add links to docs for test filtering and CI behavior manipulation

1262471

ScottTodd requested review from HereThereBeDragons and geomin12 March 30, 2026 23:42

geomin12 approved these changes Mar 31, 2026

View reviewed changes

HereThereBeDragons approved these changes Mar 31, 2026

View reviewed changes

ScottTodd added 2 commits March 31, 2026 10:01

Add comment about label suffix trimming.

0855f95

Merge remote-tracking branch 'upstream/main' into multi-arch-configure

f4acbfc

ScottTodd merged commit 947d882 into ROCm:main Mar 31, 2026
14 of 15 checks passed

ScottTodd deleted the multi-arch-configure branch March 31, 2026 17:24

github-project-automation Bot moved this from TODO to Done in TheRock Triage Mar 31, 2026

		return "\n".join(lines)


		def _non_default_callouts(ci_inputs: CIInputs, outputs: CIOutputs) -> list[str]:

Conversation

ScottTodd commented Mar 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Motivation

Technical Details

Comparison Metrics

Test Plan

Test Result

Submission Checklist

Uh oh!

HereThereBeDragons left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

HereThereBeDragons Mar 26, 2026

Choose a reason for hiding this comment

Uh oh!

ScottTodd Mar 30, 2026

Choose a reason for hiding this comment

Uh oh!

HereThereBeDragons Mar 31, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ScottTodd commented Mar 27, 2026

Uh oh!

HereThereBeDragons left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ScottTodd commented Mar 30, 2026

Uh oh!

ScottTodd commented Mar 30, 2026

Uh oh!

HereThereBeDragons left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

ScottTodd commented Mar 31, 2026

Test Plan

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

ScottTodd commented Mar 23, 2026 •

edited

Loading