Dnia 2014-09-22, o godz. 18:43:11
Diego Biurrun <[email protected]> napisał(a):

> On Sun, Sep 21, 2014 at 06:31:04PM +0200, Michał Górny wrote:
> > Dnia 2014-09-21, o godz. 11:52:25
> > Diego Biurrun <[email protected]> napisał(a):
> > > This reduces TMPDIR pollution and possibly avoids race conditions with
> > > temporary files that are not atomically created.
> > > 
> > > Based on a patch from Michał Górny <[email protected]>.
> > > ---
> > > --- a/configure
> > > +++ b/configure
> > > @@ -2593,15 +2593,18 @@ if ! check_cmd mktemp -u XXXXXX; then
> > >  
> > > +AVTMPDIR=$(mktemp -u "${TMPDIR}/avconf.XXXXXXXX" 2> /dev/null)
> > > +
> > > +mkdir $AVTMPDIR || die "Unable to create temporary directory in $TMPDIR."
> > 
> > Your patch no longer avoids race conditions since once again you are
> > using 'mktemp -u'. In fact, I don't understand at all why you are
> > creating the directory manually when 'mktemp -d' will do that for you.
> 
> There are systems without mktemp().
> 
> > Additionally, you removed the fallback for when 'mktemp' is not
> > available. You don't have any checks for that case, so configure will
> > simply attempt to run with empty $AVTMPDIR which -- depending on
> > implementation of mkdir -- will either fail because mkdir is called
> > with no arguments or continue trying to create files in root directory.
> 
> No, I'm keeping the mktemp() replacement that you removed.  Have you tested
> what happens w/o mktemp() in your implementation?  You only create the subdir
> if "mktemp -d" works, so creating the actual temporary files should fail.

Sorry, I missed that. And yes, I forgot the extra 'mkdir' in my patch.

However, I don't see any reason not to use 'mktemp -d' if 'mktemp -u'
works. Atomicity is the key point in preventing race conditions.

-- 
Best regards,
Michał Górny

Attachment: signature.asc
Description: PGP signature

_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to