Re: master 4803fba487 1/2: 'C-x v v' on a diff buffer commits it as a patch (bug#52349)

[Dmitry Gutov <dgutov@yandex.ru>]

On 31.08.2022 19:36, Eli Zaretskii wrote:
>> Date: Wed, 31 Aug 2022 19:21:12 +0300
>> Cc: Po Lu<luangruo@yahoo.com>, "Alfred M. Szmidt"<ams@gnu.org>,
>>   eliz@gnu.org,larsi@gnus.org,emacs-devel@gnu.org
>> From: Dmitry Gutov<dgutov@yandex.ru>
>>
>> On 31.08.2022 19:06, Juri Linkov wrote:
>>>> But a loop is necessary because the diff can have more than one hunk, isn't
>>>> it?
>>> A loop is necessary indeed.  The problem is that I tried it with Bzr,
>>> but it has such a bug that `bzr diff` adds an extra line at the end,
>>> so it's not (eobp) after the last diff hunk.  After removing the
>>> final newline in the*vc-diff*  buffer the loop can be finished.
>> Hmm, so how do we apply the full diff in a backend-agnostic fashion?
>>
>> Should we delegate to 'patch', then? And require it to be installed.
> Shouldn't the code which looks for hunks be resilient to such minor
> differences?  Why does it fail if there's an extra line at the end?
> Does that line look anything like another hunk?

I'm not sure how that line is supposed to look (I haven't been able to 
reproduce it with the one Bzr repo I had), but it should be possible to 
use a loop like

   (while (or (bobp) (looking-at diff-hunk-header-re)
     (diff-apply-hunk))

Anyway, I went ahead with using 'patch' for now. At least it will show 
simpler errors if user tries to commit patch with broken syntax.