Re: What's cooking in git.git (Dec 2014, #05; Mon, 29)

From: Junio C Hamano <gitster@pobox.com>
Date: Jan 7, 2015, 00:29 UTC
Message-ID: <xmqqfvbnmo0e.fsf@gitster.dls.corp.google.com>
In-Reply-To: <20150105200644.GJ29365@google.com>

Jonathan Nieder <jrnieder@gmail.com> writes:

Show 19 quoted lines

> Stefan Beller wrote:
>> On Mon, Dec 29, 2014 at 1:28 PM, Junio C Hamano <gitster@pobox.com> wrote:
>
>>> * sb/copy-fd-errno (2014-11-17) 1 commit
>>>  - copy.c: make copy_fd preserve meaningful errno
>>>
>>>  Will be rerolled as a part of a larger series.
>>
>> I am not expecting to reroll this as part of a larger series any more.
>> So if this makes sense on its own, please pick it up?
>
> I think the errno-based approach is problematic.  No callers today run
> into the problem the patch fixes and if a caller did, I fear that
> caller would be broken in other ways.
>
> That said, no one seemed too excited about the alternative that
> propagates the error message more explicitly with a strbuf.  I'd be
> happy to re-send anyway, or I can work around this patch when
> rerolling.

OK, let's resurrect the issue when more people are actively interested in having the right mechanism to propagate errors up in the callchain. I just re-checked the callers of copy_fd and I agree that this will not affect existing callers (including the one in hold_lock_file_for_append() called by add_to_alternates_file() that passes LOCK_DIE_ON_ERROR), so I am inclined to drop the patch for now.