ci: move debug to the weekly profile and trim the catchall variants - #3107
Merged
Ricardo Salveti (ricardosalveti) merged 5 commits intoSep 11, 2026
Merged
Conversation
…ance The rt-6.18-distro-kvm kernel entry appends ci/qcom-distro-kvm.yml, which includes ci/qcom-distro.yml and so replaces the distro selected earlier in the chain. For the debug and performance rows that turns 32 jobs into qcom-distro builds the qcom-distro row already produces, published under the wrong name. 798c02f excluded nodistro and qcom-distro-catchall for the same reason. Exclude debug and performance the same way. Assisted-by: Claude Code:claude-fable-5 Signed-off-by: Ricardo Salveti <ricardo.salveti@oss.qualcomm.com>
The debug jobs are 62 of the nightly's 129 machine-hours and, at 77 minutes each, the last jobs to finish, for a variant whose value is coverage of DEBUG_BUILD rather than a nightly artifact. The weekly profile exists for exactly this and currently has no rows of its own. Move debug to the weekly profile. Assisted-by: Claude Code:claude-fable-5 Signed-off-by: Ricardo Salveti <ricardo.salveti@oss.qualcomm.com>
The debugfs is a companion rootfs holding every -dbg package, and tarring it with bzip2 takes 1,600-2,000 s per image, 5,500 s on a three-image qcs9100-ride-sx build, by far the biggest task in a debug job. Switch it to zstd, which needs a fraction of the CPU for output about 1-3% larger. The published tarball is renamed from .rootfs-dbg.tar.bz2 to .rootfs-dbg.tar.zst. Assisted-by: Claude Code:claude-fable-5 Signed-off-by: Ricardo Salveti <ricardo.salveti@oss.qualcomm.com>
The three image-level SPDX tasks hash every deployed image twice, for several minutes per image, and some CI variants publish nothing that needs an SBOM. Add ci/nospdx.yml, which drops create-spdx-image-3.0 from IMAGE_CLASSES, the class that defines those tasks; package-level SPDX and its sstate are untouched. Inheriting nospdx instead does not parse: the class lives in classes-recipe, which a configuration-level INHERIT does not search. Assisted-by: Claude Code:claude-fable-5 Signed-off-by: Ricardo Salveti <ricardo.salveti@oss.qualcomm.com>
The artifacts of those variants are never fetched by a test, and the image SBOM costs 1,250 s on a debug job and 340 s on a performance job. Append ci/nospdx.yml to both rows. Assisted-by: Claude Code:claude-fable-5 Signed-off-by: Ricardo Salveti <ricardo.salveti@oss.qualcomm.com>
Contributor
Author
|
This is a move to push our most expensive runs to weekly, as I don't really think it is worth building debug and performance every day, we don't really test and use that often, and it is the most expensive part of our builds. |
Test Results 119 files + 51 715 suites +396 8h 44m 6s ⏱️ - 2h 52m 57s For more details on these failures, see this check. Results for commit 93bd74e. ± Comparison against base commit bc1ff3a. This pull request removes 2 and adds 22 tests. Note that renamed tests count towards both.♻️ This comment has been updated with latest results. |
Dmitry Baryshkov (lumag)
approved these changes
Sep 10, 2026
Koen Kooi (koenkooi)
approved these changes
Sep 10, 2026
Jose Quaresma (quaresmajose)
approved these changes
Sep 10, 2026
| profile: weekly | ||
| - name: performance | ||
| yamlfile: ':ci/qcom-distro-catchall.yml:ci/performance.yml' | ||
| yamlfile: ':ci/qcom-distro-catchall.yml:ci/performance.yml:ci/nospdx.yml' |
Contributor
There was a problem hiding this comment.
I thought we could use drop the create-spdx-image-3.0 directly on ci/performance.yml and ci/debug.yml. That way we wouldn't even need the ci/nospdx.yml.
Contributor
Author
There was a problem hiding this comment.
I think it is still valuable to have that fragment, let's keep this for now and we can revisit if needed.
Contributor
Author
|
Usual known failures. |
Ricardo Salveti (ricardosalveti)
merged commit Sep 11, 2026
6a1f836
into
qualcomm-linux:master
210 of 211 checks passed
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
5 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.
The debug rows are 62 of the nightly's 129 machine-hours and, at 77 minutes
each, the last jobs to finish. On top of that, the rt-6.18-distro-kvm kernel
entry turns the debug and performance rows into qcom-distro builds the
qcom-distro row already produces, so 32 jobs per run publish a duplicate
under the wrong name.
Drop the hijacked rt-kvm rows of debug and performance, move debug to the
weekly profile (which had no rows of its own), compress the debugfs tarball
with zstd instead of bzip2, and skip the image-level SBOM on debug and
performance through a new ci/nospdx.yml fragment.
The matrix goes from 221 to 189 configurations. Nightly and push runs lose
all debug rows; the weekly keeps them, at a fraction of the cost. The
published debug tarball is renamed from .rootfs-dbg.tar.bz2 to
.rootfs-dbg.tar.zst.