Re: [PATCH] am: add --check option

[Johannes Schindelin <Johannes.Schindelin@gmx.de>]

Hi,

On Tue, 4 Jun 2019, Johannes Sixt wrote:

> Am 04.06.19 um 00:00 schrieb Drew DeVault:
> > On Mon Jun 3, 2019 at 11:09 PM Johannes Sixt wrote:
> >> I have to wonder how --check works when 'am' applies multiple patches.
> >>
> >> When the second patch in a patch series depends on that the first patch
> >> is fully applied, what does --check do? Without the first patch applied,
> >> then a naive check of the second patch will certainly fail, doesn't it?
> >
> > Yeah, this was being discussed in another thread. It'll fail if the
> > second patch relies on changes from the first. Open to suggestions on
> > how to improve that, but I think it can be improved in a later patch.
> > One solution would be to apply all of the patches and then roll back the
> > head, but that would dirty the reflog and wouldn't work on a read-only
> > filesystem (which it ought to, imo). We can't just say bugger this for a
> > lark and ask people to use git-apply, because git-apply chokes on the
> > typical email which isn't in the one specific format git-apply wants to
> > see (git-am massages emails into that format before sending them to
> > git-apply).
> >
>
> You can 'git apply --cached' the patches on a temporary index. This
> works as long as no merge is necessary, because that would require a
> worktree.

For extra brownie points, this could be done in-memory, without writing
out any files.

Of course, while applying diffs should never need any merge, with the `-3`
option, merges might be necessary, still...

Ciao,
Johannes