Re: [PATCH] Fallback on getpwuid if envar HOME is unset

From: Jeff King <peff@peff.net>
Date: Aug 21, 2012, 19:40 UTC
Message-ID: <20120821194049.GB15667@sigill.intra.peff.net>
In-Reply-To: <7vvcgccaol.fsf@alter.siamese.dyndns.org>

On Tue, Aug 21, 2012 at 12:22:18PM -0700, Junio C Hamano wrote:

Show 22 quoted lines

> Conley Owens <cco3@android.com> writes:
> 
> > Yes, that's the upstart I'm referring to.  This makes sense.  However, it's a
> > confusing situation to run into.  Would a warning about an unset $HOME be
> > appropriate?
> 
> Unsetting HOME is an easy way to skip what is in ~/.gitconfig when
> helping other people on this list, and I wouldn't mind such a
> warning while I knowingly unset it, I can imagine other helpful
> people may find such a warning irritating and complain "I know I do
> not have $HOME set, as I earlier explicitly did unset it myself!".
> 
> So, I am on the fence on this one, but because
> 
>  (1) no warning would mean upstart scripts writers need to be aware
>      of lack of $HOME, but they need to be aware of it for reasons
>      unrelated to Git anyway; and
> 
>  (2) a warning while trying vanilla Git behaviour to help others
>      might be irritating, it is not an every day use anyway.
> 
> I do not think it matters either way in practice.

I do not use upstart, but presumably it logs the stderr of jobs it runs. Which means that a warning about unset $HOME would help debugging for people who care about looking in ~/.gitconfig, but would become a noisy nuisance for people whose jobs did not care.

Personally, I think it would be much friendlier of upstart to give the user's jobs a sane minimal environment. That is what cron has always done, setting HOME, LOGNAME, and SHELL. But that is my uninformed 5-second opinion as a long-time Unix user; I have not looked at upstart at all, so perhaps there is some argument I haven't seen.

-Peff