fix(ci): stop leaving stale versions behind in the AUR .SRCINFO - #62
Merged
Conversation
The AUR bump mirrors PKGBUILD metadata into .SRCINFO field by field, with one sed per line that embeds $pkgver. That only holds while every new source gets a sed of its own, and the icon source never did: it is matched by neither `colony-[0-9.]*::https` (the filename is `colony-icon-...`, no digit follows the dash) nor `download/v[0-9.]*/colony-linux`. So colony-bin has been published since 0.9.0 announcing `pkgver = 0.9.1` while the icon source still pointed at the v0.8.0 tag. Nothing failed. Builds come from the PKGBUILD, so the package was fine; only the metadata the AUR serves disagreed with the file next to it. Rewriting every occurrence of the outgoing version instead of naming the lines it may appear on fixes today's case, but the durable part is the check after it: if any mention of the old version survives, the release now fails and says which line. The next field somebody forgets stops being invisible. Verified by replaying the 0.8.0 -> 0.9.1 bump that produced the defect: the result is byte-identical to `makepkg --printsrcinfo` on the 0.9.1 PKGBUILD, apart from the binary checksum, which the workflow takes from the downloaded artifact rather than from the previous .SRCINFO. Generating .SRCINFO with makepkg would remove the whole class of bug, but the runner is Ubuntu and it would mean a container in the release path: an unverifiable step whose failure costs every future version bump, traded against metadata that is merely cosmetically stale. Not worth it here.
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 AUR bump mirrors PKGBUILD metadata into .SRCINFO field by field, with one
sed per line that embeds $pkgver. That only holds while every new source gets
a sed of its own, and the icon source never did: it is matched by neither
colony-[0-9.]*::https(the filename iscolony-icon-..., no digit followsthe dash) nor
download/v[0-9.]*/colony-linux.So colony-bin has been published since 0.9.0 announcing
pkgver = 0.9.1whilethe icon source still pointed at the v0.8.0 tag. Nothing failed. Builds come
from the PKGBUILD, so the package was fine; only the metadata the AUR serves
disagreed with the file next to it.
Rewriting every occurrence of the outgoing version instead of naming the lines
it may appear on fixes today's case, but the durable part is the check after
it: if any mention of the old version survives, the release now fails and says
which line. The next field somebody forgets stops being invisible.
Verified by replaying the 0.8.0 -> 0.9.1 bump that produced the defect: the
result is byte-identical to
makepkg --printsrcinfoon the 0.9.1 PKGBUILD,apart from the binary checksum, which the workflow takes from the downloaded
artifact rather than from the previous .SRCINFO.
Generating .SRCINFO with makepkg would remove the whole class of bug, but the
runner is Ubuntu and it would mean a container in the release path: an
unverifiable step whose failure costs every future version bump, traded
against metadata that is merely cosmetically stale. Not worth it here.