control: tag -1 +patch Hello,
On Fri 19 Jul 2019 at 10:55PM +01, Ian Jackson wrote: > It would be nice to check that the patches are right. (In > gbp/unapplied, that they apply and are up to date.) > > This would involve git-apply, at the very least. Running gbp pq would > add an unwanted dependency and also it is very slow. The main > reason dgit and gdr use gbp pq is to do the weird DEP-3 Debian patch > metadata thing (which should probably be abolished in Debian > completely...) - but this is not needed for a "do they apply" check. > dgit has the absurd fallback to dpkg-source patch application (git > grep "absurd") too but I think we can simply refuse to work with > packages whose patches can't be git-applied. > > For dpm and nofix it might be possible to check that the patches are > right by reverse applying them in reverse order, and expecting the > resulting tree to be identical to upstream. I've implemented this in branch 'series/git-debpush-apply-patches-v1' of repo <https://git.spwhitton.name/dgit>. Regarding the dpm and nofix check (which I'm less sure about than the gbp&unapplied check): - should we always expect to get the unmodified upstream source just by unapplying patches? - should the two calls to check_fail in check_patches_unapply be distinct named checks? - needs a proper test case. I have ad hoc tested with nofix. The four tagupl tests pass if you hack the full path to git-playtree-create into the script -- it seems that using-intree has not been taught to make git-playtree-create available to git-debpush in the right way. -- Sean Whitton
signature.asc
Description: PGP signature