Please review the logic in the graph below. Does it cover all cases?
** EDIT 2022-11-11: Luca found pkgver=9999 case missing, added it
Contains all commits from !3608 (merged) and still has a bug (see todos below), code review can be done later
Description:
Update the script once more with the goal that we no longer need to
comment on pkgrel/pkgver bumps. Parse not only the version of the old
and new package, but also the checksums and remote sources.
Replace the previous check of simply having $pkgver-r$pkgrel higher
than the upstream APKBUILD with more detailed rules:
APKBUILD has remote sources? ------> Did sha512sums change? | No | | | Yes | No | Yes v | v Did the file name of a remote | * Increase pkgver source change? ----. | * Set pkgrel=0 | | | | Yes | No | v v v Is pkgver=9999? -----> * Don't change pkgver (Forked from Alpine) Yes * Bump pkgrel by 1 | | No v * Set the pkgver to the version of the source tarball * Set pkgrel=0
Besides the new rules, print a detailed error message below each package
where the test failed that explains exactly what was expected and what
the reasoning behind that is. Also add a hint that users can run this
locally with 'pmbootstrap ci' now.
Use logging instead of print, so we get colored ERROR: and NOTE:
statements from pmbootstrap code, this should make it easier to read.
TODO:
extend the graph: maintainer change (needs pkgrel bump as metadata changed), _pmb_recommends change (no pkgrel bump, pmbootstrap just reads the APKBUILD)
implement these in the check
FIXME: to check whether a file is remote or not, a glob must be used since the filenames mentioned in sha512sums may be in a subdir
