Re: [PATCH/RFC 2/2] change all unchecked calls to setenv to xsetenv

From: Junio C Hamano <gitster@pobox.com>
Date: Dec 13, 2011, 18:33 UTC
Message-ID: <7v1us8wccm.fsf@alter.siamese.dyndns.org>
In-Reply-To: <20111213181946.GC1663@sigill.intra.peff.net>

Jeff King <peff@peff.net> writes:

Show 16 quoted lines

> On Tue, Dec 13, 2011 at 01:10:27PM +0100, Erik Faye-Lund wrote:
>
>> While reviewing some patches for Git for Windows, I realized that
>> we almost never check the return-value from setenv. This can lead
>> to quite surprising errors in unusual sitations. Mostly, an error
>> would probably be preferred. So here we go.
>> 
>> However, I'm not at all convinced myself that all of these make
>> sense; in particular settings like GIT_EDITOR and GIT_PAGER could
>> perhaps benefit from having a warning printed rather than a hard
>> error.
>> 
>> Thoughts?
>
> I wrote almost the same patch once[1], but failed to actually push it
> through to acceptance.

In both cases, the patches are _designed_ to fail to attract any attention. Your earlier one had this preamble:

  Here is a patch. I still feel a little silly writing this. The chances
  that you will run out of memory doing setenv but _not_ doing any of the
  other git operations seems very low.

which rather *loudly* says "ignore me, please!" ;-)

Erik's patch this round is no better; if its log message said something like "On platform X, the environment space is merely nKB and setenv has much higher chance of failing than on typical Linux boxes", it would have been a no brainer for me to respond with "makes perfect sense but don't we also use putenv?" immediately.