Re: intended MBF: wrong redirections in maintainer scripts

[Ralf Treinen <treinen@free.fr>]

Hello Wouter,

On Tue, Aug 07, 2018 at 12:38:32PM +0200, Wouter Verhelst wrote:
> On Sat, Aug 04, 2018 at 01:15:57PM +0800, Ralf Treinen wrote:
> > Hi,
> > 
> > as announced in our talk at debconf'18 [1] we intend a MBF about wrong
> > redirections in maintainer scripts. In general these are of the form
> > 
> >   foo 2>&1 1> /dev/null
> > 
> > Here it was probably intended to send both stderr and stdout to /dev/null.
> 
> What makes you say that? ;-)
> 
> It may be that the maintainer did indeed want stdout to be discarded,
> but stderr not; for instance because they wanted to parse the stderr
> output.
> 
> (not saying this is the most likely case, but you might want to
> double-check that before filing the bugs)

We were assuming that the author of the script wanted to send both 1 and 2
to /dev/null and was victim to the common mistake of getting the order of
redirections wrong.

You are absolutely right that our assumption about the authors intention
may be wrong, and that they really intended the redirection the way they
wrote it. This means that we should be more careful, and ignore cases in
which the above pattern occurs in a context which has itself a redirection.
As a consequence, we wouldn't detect a bug in a case like this one:

 (foo 2>&1 1> /dev/null) | /some/processing

This should be enough to eliminate false positives, right?

-Ralf.