fix: repair the 0.1.21 rpm failure, and stop the two workflows racing - #36
Merged
Spacefreak18 merged 3 commits intoSep 9, 2026
Merged
Conversation
My packaging change made the clone in %prep conditional on $RPM_SOURCE_DIR/simapi being absent, and dropped the `rm -rf` that used to remove it. On a GitHub-hosted runner that is harmless -- the directory never exists -- but the fedora-43 job runs on a self-hosted runner where ~/rpmbuild/SOURCES persists between builds, so the clone was skipped and whatever tree was left over from the previous release got built instead. Staging is now opt-in via --define "staged 1". Without it the behaviour is exactly what it was before: remove any previous checkout, clone fresh. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
%prep now clones unless --define "staged 1" is passed, so the job that stages the checkout has to say so. Without this the rpm would be built from a fresh clone of the default branch rather than the commit being released. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
These two jobs and the ones in packages.yml build the same artifacts and upload them to the same release under the same filenames -- libsimapi- fedora-43.rpm, simd-debian-stable.deb and the rest. On a tag they race. That is my fault: I added packages.yml without noticing this workflow already released those names. They are also the last self-hosted jobs here. build-simapi-rpms is what failed on 0.1.21, with simd.c:12 unable to find yder.h: the job installs no dependencies at all, relying on whatever was hand-installed on that machine, and Fedora packages neither yder nor orcania. Three months after the last release the runner no longer had them. packages.yml builds fedora-43 and 44 in containers and vendors both as static libraries, so the same rpm is produced on an ordinary GitHub-hosted runner with nothing to maintain by hand. It also adds fedora-44, which was never covered here. The deb matrix is the same story: ubuntu-latest, debian-latest and debian-stable are runner labels resolving to self-hosted machines, and packages.yml builds all three in containers. The nix jobs stay -- they run on ubuntu-latest and nothing else covers them. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Spacefreak18
pushed a commit
that referenced
this pull request
Sep 10, 2026
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
2 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.
0.1.21failedbuild-simapi-rpms (fedora-43)withsimd.c:12:10: fatal error: yder.h: No such file or directory. 0.1.18, 0.1.19 and 0.1.20 all passed, so this is a regression from my packaging PR. Sorry — two separate mistakes, both mine.1. %prep stopped starting from a clean checkout
I made the clone conditional and dropped the
rm -rf $RPM_SOURCE_DIR/simapithat preceded it:On a GitHub-hosted runner that is harmless because the directory never exists. The fedora-43 job runs on a self-hosted runner where
~/rpmbuild/SOURCESpersists between builds, so the clone was skipped and a stale tree was built. I tested this only on hosted runners, where the bug is invisible.Staging is now opt-in via
--define "staged 1". Without it the behaviour is exactly as before: remove the old checkout, clone fresh.2. The two workflows upload the same filenames
packages.ymlandci.yamlbuild the same artifacts and attach them to the same release aslibsimapi-fedora-43.rpm,simd-debian-stable.deband so on — identical names. On a tag they race. I addedpackages.ymlwithout noticing this workflow already released those names.This drops
build-simapiandbuild-simapi-rpms, keeping the two nix jobs.packages.ymlcovers everything they did, in containers onubuntu-latest:It also removes the dependency that broke 0.1.21.
build-simapi-rpmsinstalls nothing — it creates rpmbuild dirs, curls the spec and runsrpmbuild, relying entirely on what was hand-installed on that machine. Fedora packages neither yder nor orcania, so those must have been placed there manually and were gone three months later.packages.ymlvendors both as static libraries, so there is nothing to maintain by hand and nothing to silently rot.Drop the second commit if you would rather keep those jobs — the first two fix the failure on their own, though the filename collision would remain.
🤖 Generated with Claude Code