On Tue, 16 Jun 2026 at 08:54, Daniel P. Berrangé <[email protected]> wrote: > > On Mon, Jun 15, 2026 at 01:17:23PM -0700, Pierrick Bouvier wrote: > > We add a new script: scripts/check-maintainers-file.py, that will run at > > configuration time (and not at build time), to not hurt build time. > > This script runs in 0.2s on my dev VM, which has an old cpu. > > > > We can expect things to be mostly in sync since adding or removing a > > source or test file will trigger a configure step. > > For the rest, like docs, tcg tests, or remaining files, GitLab CI will > > build things from scratch and always run the configure step. > > > > With this, it should be impossible by design to have an upstream > > MAINTAINERS file with non existing file entries. > > Accuracy of the MAINTAINERS file is an upstream-only concern, but > IIUC, this check is going to apply universally to every build of > QEMU which is undesirable. It is irrelevant to end users and not > appropriate to check in downsteam vendors forks. > > In a "normal" modern project this kind of check would be done in > a CI job on the merge request, since that's the only place it is > relevant. In our case, the nearest fit is the checkpatch.pl file.
You could put it in "make check", or (if you want to avoid the downstream-forks issue) in some sub-bit of "make check" that we don't check by default but do check in one of our CI jobs. checkpatch isn't a great place for checks that you want to stay consistent on because so much of its output is "this is wrong/merely something to check, ignore". -- PMM
