Sean Whitton writes ("Bug#1111548: tag2upload service makes .origs with
un-defused gitattributes"):
> git-deborig(1) says
>
> git-deborig will override gitattributes(5) that would cause the
> contents of the tarball generated by git-archive(1) not to be
> identical with the commitish archived: the export-subst and
> export-ignore attributes.
>
> So the problem is that there are git attributes which cause this that it
> doesn't defuse, atm. It wouldn't be a breaking change to its declared
> API to make git-deborig disable these other attributes, too.
Thanks for investigating.
There is a test in the dgit test suite which can sometimes detect when
new git versions introduce new hazardous attribnutes.
I suggest adding the git-archive-affecting attributes to both
git-deborig and dgit setup-gitattributes (changing its behaviour,
which was always a bit odd). Maybe the list of attribute names could
be moved to Core.pm.
The test case should be extended so that it feeds all the pathological
test caes to git archive too. Probably it should be reviewed more
generally it to check that it is fierce enough.
Note that this test does sometimes block src:git from migrating in
Debian. In 2017 I filed a bug against src:git asking for a way to
defuse even all future attributes, #851679. I'm pretty sure it has
blocked new git on at least one and possibly two occasions, until dgit
was updated to defuse those new gitattributes too.
Ian.
--
Ian Jackson <[email protected]> These opinions are my own.
Pronouns: they/he. If I emailed you from @fyvzl.net or @evade.org.uk,
that is a private address which bypasses my fierce spamfilter.
