[amd-staging-rocgdb-18] ROCgdb cherry picks from origin/amd-staging (2026-08-14) - #281
Draft
lumachad wants to merge 12 commits into
Draft
Conversation
Remove secrets: inherit from all reusable workflow calls since no named secrets are explicitly referenced; AWS authentication is handled via OIDC. Remove the /runner/config bind-mount and AWS_SHARED_CREDENTIALS_FILE env var from the build container, as boto3 will use credentials set by the OIDC configure-aws-credentials step instead. Guard the Post Build Upload step against fork PRs to prevent S3 uploads from untrusted forks. Addresses ROCM-26823. Co-Authored-By: Claude <noreply@anthropic.com> (cherry picked from commit f5601c3)
TheRock CI and pre-commit were running on both push and pull_request events. Since every change goes through a PR, the push runs on amd-staging are redundant and waste CI resources. Co-Authored-By: Claude <noreply@anthropic.com> (cherry picked from commit 0e4734a)
(cherry picked from commit a58f9de)
…efully Issue an XFAIL only when the bug presents itself, avoiding an XPASS. (cherry picked from commit 8d57cb8)
Previously the build job hardcoded azure-linux-scale-rocm (Azure) and the CPU test job hardcoded aws-linux-scale-rocm-prod (AWS). Both labels are now selected dynamically from ROCm/therock-ci-config's runner-config.json (build_runners.linux.default, highest-weight entry) so CI tracks fleet changes without manual updates. The setup job in therock-ci.yml checks out therock-ci-config and reads the runner label via ci_config_api.py, passing it as build_runs_on to therock-ci-linux.yml (new workflow_call input). therock-test-packages.yml applies the same logic in configure_test_matrix for the CPU test runner. Both steps fall back to aws-linux-scale-rocm-prod if therock-ci-config is unavailable (e.g., in forks). The runner-selection logic is extracted into .github/scripts/select_runner.py to avoid duplication between the two workflows. Co-Authored-By: Claude <noreply@anthropic.com> (cherry picked from commit cc4697e)
Move THEROCK_COMMIT_REF and the build container digest out of therock-ci-linux.yml and into .github/configs.json, a plain data file documented in .github/configs.md. A resolve job reads and validates the pins and exposes them as outputs; the build job references those outputs for container.image and THEROCK_COMMIT_REF. This lets the automated update workflow push changes with only contents: write on the GITHUB_TOKEN. The update script is reworked accordingly: - Fetch the TheRock branch HEAD and committer timestamp from the GitHub REST API (retried on transient failures) instead of git ls-remote and date.today(). Fetch the build image creation timestamp from the OCI config blob instead of date.today(). Both timestamps are stored in full ISO 8601 format and gated per pin so an unchanged value keeps its original timestamp. Timestamp fetches are skipped in dry-run mode. - Validate all API response fields with descriptive errors; guard required config keys before any git side-effects. - PR title is "Automated TheRock dependency update (<date>)"; body lists each bumped pin with its short ref and date. - Print both tracked values on every run with clear update / unchanged labels; name what would change in dry-run mode. Add a comprehensive unit test suite covering all public functions, including retry logic, error paths, timestamp gating, and the run_update state machine. (cherry picked from commit 1dfad84)
The maint print address-spaces command previously printed only a flat list of address-space names. Rewrite address_spaces_dump to query dbgapi for five properties per address space and emit them as a structured ui_out table: name, DWARF number, address size in bits, null address, and access class. A new helper, address_space_access_to_string, maps the amd_dbgapi_address_space_access_t enum to a human-readable string (all, program-constant, dispatch-constant). The command help text and the maint print address-spaces section in gdb.texinfo are updated to document the new columns. A new test, gdb.rocm/maint-print-address-spaces.exp, runs the command, and checks that a header row and at least one data row are printed without error. (cherry picked from commit 92775a3)
Add -f/--fraction F as an alternative to -j/--jobs for controlling parallelism. The two options are mutually exclusive. When --fraction is given, the job count is resolved as max(1, int(cpu_count * F)), allowing callers to request a hardware-proportional limit without knowing the exact core count of the target machine. Both options still require --parallel. (cherry picked from commit 33f3678)
Move the inline python3 -c validation out of the resolve step into a standalone .github/scripts/config.py script. The script validates each requested field against an allow-list of known fields and their expected formats (40-char hex SHA for therock_commit_ref, ghcr.io sha256 digest for build_image), and writes field=value pairs to $GITHUB_OUTPUT. Also guard the validate() call against non-string field values (e.g. null or a number in configs.json) so the script emits the friendly error message rather than an uncaught TypeError traceback. (cherry picked from commit e761ce9)
…rors.exp
Running gdb.rocm/hip-catch-errors.exp in an environment where the
__hipOnError symbol is not defined gives
ERROR: tcl error sourcing <path-to>/gdb/testsuite/gdb.rocm/hip-catch-errors.exp.
ERROR: can't read "res": no such variable
while executing
"return $res"
("uplevel" body line 27)
invoked from within
"uplevel 1 $body"
invoked from within
"with_test_prefix preparation_test {
with_rocm_gpu_lock {
clean_restart ${::testfile}.dbg
if {![runto_main]} {
return false
}
gdb_test_mul..."
Fix this by defining "res" at an earlier point.
The problem here is in fact deeper. The code body that is used in
preparation_test has return statements. One would expect the return
statements to actually return from preparation_test early and not even
reach `return $res`, but that's not happening because of how
with_rocm_gpu_lock and with_lock are defined and use Tcl's uplevel.
Fixing them will be more involved and should be addressed separately.
(cherry picked from commit 3a93be6)
A system-wide ROCM_PATH can redirect LLVM tool lookups away from the artifacts tree configured by setup_environment, causing hard-to-diagnose test failures. Strip it from the subprocess environment whenever the caller has not explicitly opted in via --try-rocm-path. (cherry picked from commit 128d02e)
github-actions
Bot
changed the title
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.
Commits:
03249c8 github: drop ROCM_PATH when --try-rocm-path is not passed
a8af09d gdb, testsuite: fix 'no such variable' error in gdb.rocm/hip-catch-errors.exp
3f88d0a github: improve TheRock dependency update workflow
1da6a00 github: test_rocgdb.py: add --fraction/-f parallel job scaling
df2a43f gdb: extend maint print address-spaces to show all dbgapi available info
28717a4 Revert "github: improve TheRock dependency update workflow"
f6b393d github: improve TheRock dependency update workflow
4280ac3 [ci] Fetch build and CPU test runner dynamically from therock-ci-config
54fb5dd gdb.rocm/omp-target-basic: handle firstprivate compiler type bug gracefully
7c39ce7 gdb.rocm/alu-exceptions: prevent ROCr core dumps on exceptions
b4f6bab github: remove push triggers, run CI only on PRs
c126c43 github: tighten CI workflow security settings