Re: [PATCH] mc crashes when temporary directory cannot be created

[Jindrich Novy <jnovy redhat com>]

Hi Pavel,

On Tue, 2007-01-02 at 16:49 +0200, Pavel Tsekov wrote:
> On Tue, 2 Jan 2007, Jindrich Novy wrote:
> > On Tue, 2007-01-02 at 16:28 +0200, Pavel Tsekov wrote:
> >> On Tue, 2 Jan 2007, Jindrich Novy wrote:
> >>> On Sat, 2006-12-30 at 20:51 +0200, Pavel Tsekov wrote:
> >>>>> There is an even simpler cure. In mc_tmpdir() when executing
> >>>>> the fallback code pass an absolute path to mc_mkstemps().
> >>>>> This will prevent the loop. However I am not yet conviced
> >>>>> that this is the best solution. I am still investigating.
> >>>>
> >>>> I am attaching a patch which passes an absolute path to mc_mkstemps()
> >>>> when invoked from mc_tmpdir(). What do you think about this fix ?
> >>>> I may add a comment why it is necessary to call mc_mkstemps() with
> >>>> an absolute path. By the way I think we should add a check whether
> >>>> the environment variable TMPDIR contains an absolute path. Anyway, I'll
> >>>> leave this for another patch.
> >>>
> >>> Thanks for the patch. Do you plan to convert the relative paths to
> >>> absolute ones when detected?
> >>
> >> I think TMPDIR should be ignored in this case. What's your opinion ?
> >
> > So some hardcoded value such as "/tmp" will then be assumed in case of
> > relative path in TMPDIR?
> 
> Btw. does it make sense to use relative TMPDIR ?!

This was the question I asked a few people around in person, the
conclusion was that they could imagine a situation when they would use
relative TMPDIR, but personally I consider that pretty ugly. So the
fallback to a hardcoded value seems to fit here in case of relative path
in TMPDIR as it is pretty unlikely to be seen anywhere.

Jindrich