[llvm-dev] Phabricator -> GitHub PRs?

[Renato Golin via llvm-dev]

On Wed, 15 Jan 2020 at 18:55, Nicolai Hähnle <nhaehnle at gmail.com> wrote:
> This has simply not been true in my experience. Actually, not having
> to re-send a new series is one of the main advantages that
> Phabricator-based review has over the original review style for Git,
> which is to send patch series via mailing lists.

Interesting. If they can be committed separately, why were they a
series in the first place?

Or was that independent, but related, patches? In this case, they
could have been different PRs with mention of each other, no?


> It might be the case that you occasionally have series where major
> redesigns are required, and then asking for a fresh start makes sense.

What I mean by patch series is exactly what you would have in a
sanitised git branch: a number of sequential patches that build on
each other. You cannot commit patch 3 before 2 as Git wouldn't let you
merge.

In those cases, if we want to have patch 3 before the series, the
author needs to rewrite history (rebase -i), push the patch up and
pull into another branch, so that we can have the old tree with -1
patches as the derived series, and the cherry-picked patch merged
sooner.

But in those cases, the patch would most certainly have to be
different, and that would also change the remaining series. So, new
patch, new series.

Makes sense?

--renato