Re: [PATCH] git-am: Handle "git show" output correctly

From: Junio C Hamano <gitster@pobox.com>
Date: Sep 12, 2012, 22:19 UTC
Message-ID: <7vhar2c29s.fsf@alter.siamese.dyndns.org>
In-Reply-To: <CAPBPrntHOpDaH3cSNiKKxVJDbZTpPrLc99TgPM_GdoXecUvs9w@mail.gmail.com>

Dan Johnson <computerdruid@gmail.com> writes:

Show 8 quoted lines

>> Not really.  If we start encouraging people to use "git show" output
>> as a kosher input to "am", we would have to support such use
>> forever, and we end up painting ourselves in a corner we cannot get
>> out of easily.
>
> If git am emitted a warning when accepting "git show" output, it seems
> like it would support Peter's use-case without encouraging bad
> behavior?

Are you seriously suggesting me to sell to our users a new feature saying "this does not work reliably, we would not recommend using it, no, really, don't trust it." from the day the feature is introduced, especially when we know it will not be "the feature does not work well yet, but it will, we promise" but is "and it may become worse in the future"?

I do not see much point in doing that.

Besides, what bad behaviour do we avoid from encouraging with such an approach? As Peter said, the problem is not on the part of the user who ended up with an output from "git show", when he really wants output from "git format-patch". Giving the warning to the user of "git am" is too late.

I may be able to be pursuaded to swallow a new script somewhere in the contrib/ hierarchy that takes a "git show" output and formats it to look like "format-patch" output to be fed to "git am". That way, when a user has trouble with its parsing of "git show" output, at least we can ask for the output of the format massaging step to help us diagnose where the problem lies.