Re: severity for bugs in ignoring TMP/TMPDIR?
On Mon, Feb 06, 2012 at 03:59:07PM +0100, Bernd Zeimetz wrote:
> On 02/05/2012 12:22 PM, Alastair McKinstry wrote:
> > On 2012-02-05 11:04, Sune Vuorela wrote:
> >> On 2012-02-05, Paul Wise <pabs@debian.org> wrote:
> >>> If I notice that software in Debian is ignoring TMP/TMPDIR (since I use
> >>> libpam-tmpdir), what severity should I file the resulting bugs at?
> >> wishlist?
> >>
> >> /Sune
> >>
> > Depends on how bit the files it uses in tmpdir. I've a bug in some code
> > i'm using
> > ( a FUSE filesystem that uses a cache in /tmp) that runs as root, that
> > at times places
> > arbitrarily large files (in my workflow, 5-100 GB) in /tmp, which is on
> > the root filesystem.
>
> Well, probably you might want to move /tmp away from your root fs...
>
> > When a process accesses a 100 GB file it overflows / . So I needed to
> > get it to use an alternate $TMPDIR.
> > Ignoring $TMPDIR is a critical severity bug for me.
>
> ... and then it is just wishlist again. Having /tmp on / is as bad as
> having /var/log on / on busy machines.
And how about having /tmp in tmpfs, so that it is limited to a
fraction of the size of memory? No application should assume that it
can store working files of arbitrary size in /tmp.
Ben.
--
Ben Hutchings
We get into the habit of living before acquiring the habit of thinking.
- Albert Camus
Reply to: