Re: [Bug-tar] avoid spurious test failure (working --listed), #24

[Ralph Corderoy]

Hi Sergey,

> Ralph Corderoy <address@hidden> wrote:
> > The PATH shows the commercial product CoSort is installed.  This can
> > replace AIX's normal sort.  It's intended to be faster.
> 
> Ah, thanks for the explanation.
> 
> > I suspect it spews stats to stderr, possibly switchable.  The user
> > needs to improve his AIX environment, at least when building
> > external software.
> 
> Well, I guess so. He reported later that using native sort solved the
> problem, yet I think we'd better be prepared for such cases in the
> future.

No, I'm sorry, but I disagree.  sort's stderr is silent unless there's a
problem.  There's no expected problem so stderr shouldn't be discarded;
it weakens the test.  Not discarding may cause problems on the < 0.01%
of platforms where sort spews crap on stderr but that shows a platform
problem, not a tar one, and it would bite more than just tar so tar
shouldn't work around it.  More valuable would be the stderr output
where there is a genuine problem that needs tackling, but that would be
lost.

It's a slippery slope down which lies coping with other odd-ball
behaviour by one-man-and-his-grep.

Cheers,


Ralph.