Sean Whitton writes ("Bug#932476: git-debpush gbp/unapplied upstream tree
check"):
> Patch attached.
>
> Based on series in #932459.
Thanks.
> +# ---- Upstream tag tree nonidentical
> +
> +if ! [ "x$upstream_tag" = "x" ] \
> + && ([ "$quilt_mode" = "gbp" ] \
> + || [ "$quilt_mode" = "unapplied" ]) \
> + && git diff-tree --numstat "$upstream_tag" "$branch" \
> + | grep -qvP "[0-9]+\t[0-9]+\tdebian/"; then
> + fail_check upstream-nonidentical \
> + "the upstream source in tag $upstream_tag is not identical to the upstream
> source in $branch"
> +fi
I'm afraid I don't like this very much.
* Can we lead with a case which switches on the quilt mode ?
That will provide a place to add other checks for other quilt modes.
* Please use the git file selector syntax and git-diff --exit-code
rather than grepping the output of git diff-tree --numstat.
You will want to be careful to distinguish exit status 1 meaning
"different" from other failures, to avoid something like
git: diff: arrgh, my innards have exploded!
git-debpush: the upstream source in tag blah is not identical ...
So set +e, rc=$?; set -e.
* I think in gbp and unapplied quilt modes, $upstream_tag will always
be set by now (if not, then this code is too early), so don't
check for that.
* In gbp mode you must tolerate .gitignore differences.
Sorry,
Ian.
--
Ian Jackson <ijack...@chiark.greenend.org.uk> 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.
