Bug#814997: Reverting upstream commit 40ed879db fixes gnome-pkg-tools

[Andreas Henriksson]

Hello Santiago!

On Thu, Feb 18, 2016 at 11:32:29AM +0100, Santiago Ruano Rincón wrote:
> Hi,
> 
> El 17/02/16 a las 20:51, Andreas Henriksson escribió:
> > Control: reassign -1 gnome-pkg-tools
> > 
> > Hello again.
> > 
> > On Wed, Feb 17, 2016 at 07:56:29PM +0100, santiagorr at riseup.net wrote:
> > [...]
> > > grep is less and less tolerant against invalid unicode characters.
> > > Maybe are there invalid characters in the debian/* files where
> > > uploaders.mk get the info to fill the Uploaders field? 
> > [...]
> > 
> > You might actually be right here. (I was only thinking about the data
> > used as an argument for grep - not the standard input data.)
> > 
> > $ iconv -f utf-8 -t ascii < /usr/share/gnome-pkg-tools/pkg-gnome.team  | tail -n 2
> > iconv: illegal input sequence at position 713
> > Loic Minier <lool at dooz.org>,
> > 
> 
> Try iconv -f UTF-8 -t ASCII//TRANSLIT /usr/share/gnome-pkg-tools/pkg-gnome.team

The file itself seems to be properly formatted as far as I can see and
grep foo < /usr/share/gnome-pkg-tools/pkg-gnome.team work as well.
The problem is probably introduced in the uploaders.mk makefile mangling
which uses echo, sed, etc to create the standard input data for grep.
Not sure how to make a reproducible testcase still though.
(I tested ordering the sed of pkg-gnome.team before the echo just
in case the starting BOM would be the issue but no change.)


> 
> > And indeed, dropping Loïc and Sebastian Dröge from the list fixes the
> > problem.
> > 
> 
> Please, don't drop these folks just by their non-ascii compliant names!

It was just a test, but also the list is long overdue for a cleanup.
Noone ever seems to remove their name when they no longer are
active in pkg-gnome maintenance.

> 
> Actually, my up-to-dated pbuilder is able to build totem, under my
> default es_CO.UTF-8 locale. And I am able to grep their names:
> 
> % grep -F -e ï -e ö /usr/share/gnome-pkg-tools/pkg-gnome.team
> Loïc Minier <lool at dooz.org>,
> Loïc Minier <lool at debian.org>,
> Sebastian Dröge <slomo at debian.org>,
> 
> but under an ascii context:
> 
> % LANG=C grep -F -e ï -e ö /usr/share/gnome-pkg-tools/pkg-gnome.team
> Binary file /usr/share/gnome-pkg-tools/pkg-gnome.team matches
> 
> comparing against grep-2.22:
> 
> % LANG=C grep -F -e ï -e ö /usr/share/gnome-pkg-tools/pkg-gnome.team
> Loïc Minier <lool at dooz.org>,
> Loïc Minier <lool at debian.org>,
> Sebastian Dröge <slomo at debian.org>,
> 
> 
> Maybe this is actually an issue in grep, or an encoding not matching
> between the input file and the builders's locale.
> What locale are you using? From your message filling the bug report:
> 
> Locale: LANG=C, LC_CTYPE=C (charmap=ANSI_X3.4-1968) 
> 
> So maybe that is the problem.

I'll try dropping my pbuilder chroot tarball and recreating it.

> 
> See https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=800670 
> and upstream's comment at http://debbugs.gnu.org/cgi/bugreport.cgi?bug=19230#8
> 
> 
> A possible workaround with grep-2.23 is to force it to process the file
> as text, using the -a option, even if it is not a binarly file:
> 
> % LANG=C grep -aF -e ï -e ö /usr/share/gnome-pkg-tools/pkg-gnome.team
> Loïc Minier <lool at dooz.org>,
> Loïc Minier <lool at debian.org>,
> Sebastian Dröge <slomo at debian.org>,

Thanks for your followup and this workaround which sounds like it could
be a good idea anyhow for uploaders.mk!


Regards,
Andreas Henriksson