"Kristoffer Haugsbakk" <code@xxxxxxxxxxxxxxx> writes:
> On Thu, Mar 23, 2023, at 17:22, Oswald Buddenhagen wrote:
>> Instead of generating a silly-looking `Revert "Revert "foo""`, make it
>> `Reapply "foo"`.
>
> Nice addition.
>
> $ echo change >> README.md
> $ ./bin-wrappers/git add README.md
> $ ./bin-wrappers/git commit -m 'A change'
> $ ./bin-wrappers/git revert --no-edit @
> $ ./bin-wrappers/git revert --no-edit @
> $ ./bin-wrappers/git log --oneline master..
> adfce56c6a (HEAD -> reapply) Reapply "A change"
> 395894c2ce Revert "A change"
> a01e3d6b3d A change
> 058643b69f sequencer: beautify subject of reverts of reverts
I think Oswald saw the end of a thread a few years ago that
discussed a similar idea and the patch is a continuation of that
thread, but what should happen when the re-application turns out to
be bad?
The significance of the act of reverting such a reapplication to the
project would be different from the initial revert (where "yikes,
the change introduces a regression---let's revert to the state
before the change, regroup, and see what we should do" was the
motivation). Somebody thought that it now was OK to reintroduce the
change, presumably because the code paths around it now have become
ready for it and the phrase "Reapply" makes a perfect sense to
describe it ("Revert an earlier revert" is not too bad, either,
though). But then it turns out it still was a bad idea. Should we
say
Revert Reapply "A change"
Next time somebody thinks the code paths around there are finally
ready, do they do
Reapply Reapply "A change"
or something else? That may be shorter, but even more cryptic than
'Revert Revert Revert Revert "A change"'---"Revert^4 "A change"" the
other thread proposed start to look less horrible. So, I dunno.
