diff options
author | Ian Jackson <ijackson@chiark.greenend.org.uk> | 2018-08-25 01:05:22 +0100 |
---|---|---|
committer | Ian Jackson <ijackson@chiark.greenend.org.uk> | 2018-08-25 12:26:51 +0100 |
commit | 2da6bc9c434e12dd33e4506c3eaf605b87ec30d8 (patch) | |
tree | d61f8e68a06f99a910ef989ed18ef2c2b6113d21 /tests/Makefile | |
parent | d7f555bf5c00a135d3991e9829450c4de73cafaa (diff) |
dgit: Replace branch_is_gdr with a history walker
The debrebase-last approach is fundamentally wrong, because, for
example, cloning a stitched laundered gdr branch from a git server
does not establish debrebase-last, so dgit would make patches
itself (slowly).
So, instead, use git-debrebase make-patches in situations where it
will succeed. We could just run it but that's slower, and it is a bit
awkward because we have to consider whether gdr is installed. So
provide our own implementation. It can be simpler because it only has
to handle the easy cases;
On testing of the new algorithm: we want to be confident that this
doesn't misfire; otherwise dgit users could be inconvenienced. And we
want it to work for gdr users of course.
We can analyse the test coverage of the logic in branch_is_gdr
by running the whole test suite and then comparing this:
cat tests/tmp/*.log |perl -ne 'next unless s/^branch_is_gdr \w+ //; print' |sort -u |less
with this:
git-grep 'branch_is_gdr ' | perl -pe 's/^^dgit:\s+//' |sort |less
Noting that we should ideally have one each of all the gdr kinds we
try to recognise. Currently that shows that we are missing only:
gdr merged-breakwater
unmarked BreakwaterStart YES
Closes: #907208.
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
Diffstat (limited to 'tests/Makefile')
0 files changed, 0 insertions, 0 deletions