Sam Hartman writes ("Bug#928554: dgit-nmu-simple should give an example of
generating a patches-unapplied nmudiff"):
> The dgit-nmu-simple man page doesn't give any explicit examples of
> generating an nmudiff.
Thanks for bringing this up. See also #850560 "Want `dgit nmudiff`".
> It's unclear that we actually have a standard on whether an nmudiff
> should be patches-applied or patches-unapplied.
> I find that I typically get patches-unapplied nmudiffs, that if you use
> a standard dpkg workflow without dgit they are easier, and having the
> patch duplicated is harder to review.
There are at least *three* possible formats:
* git diff dgit/dgit/sid..dgit/sid # after quilt fixup
diff of whole package, patches applied
upstream changes present as diff and as interdiff in debian/patches
rune will work fine on non-quilt package
* git-format-patch dgit/dgit/sid # before quilt fixup [1]
(or git diff dgit/dgit/sid..dgit/sid # before quilt fixup [1])
diff of whole package, no patches generated
upstream changes present as diff to upstream files, only
diff contains no changes to debian/patches
rune will work fine on non-quilt package
if git-format-patch is used, maintainer can git-am in their
gbp pq patch queue branch or equivalent.
* git diff dgit/dgit/sid..dgit/sid :!/ :/debian # after quilt fixup
diff of only packaging including any patches
upstream changes present as interdiff in debian/patches, only
rune will NOT work on non-quilt package
rune will NOT work for new upstream version
[1] Or with appropriate restrictions to show only "real" changes.
Maybe we should do whatever debdiff does.
> I find it's unintuitive how exactly you generate a good
> patches-unapplied nmudiff, so I looked for guidance in the man page and
> didn't find any.
Mmm. My 3rd rune above will do this, I think. (Untested.)
Ian.
--
Ian Jackson <[email protected]> These opinions are my own.
If I emailed you from an address @fyvzl.net or @evade.org.uk, that is
a private address which bypasses my fierce spamfilter.
