reopen 623578
tags 623578 - patch
clone 623578 -1
notfixed -1 1.16.7
reassign 623578 gadfly
found 623578 1.0.0-15
block 623578 by -1
thanks
Hi!
On Sun, 2012-10-14 at 23:20:11 +0200, Samuel Thibault wrote:
> Andreas Metzler, le Sun 14 Oct 2012 15:44:16 +0200, a écrit :
> > Afaict (Both by looking at 684625 which is not fixed and by checking
> > the newest binNMU on my up to date testing system) we are
> > currently still doing binNMUs with something equivalent to this:
> >
> > dch --bin-nmu foo && dpkg-buildpackage -B
> >
> > Which works perfectly fine with gadfly:
> > -rw-r--r-- 64708 Oct 14 08:53 python-kjbuckets_1.0.0-15+b1_i386.deb
> > -rw-rw-r-- 970 Oct 14 08:53 gadfly_1.0.0-15+b1_i386.changes
> >
> > I get the same result if I add ', binary-only=yes' to
> > debian/changelog. Please describe how the issue can be reproduced?
In this case using binary-only=yes would produce a correct Source
field (with the right source version), something currently not
happening w/o it.
> I can't reproduce any more with Wheezy apparently, only in Squeeze.
> My guess is that it's dpkg 1.16.7 which fixed the issue, marking as
> such.
That's because that dpkg version is buggy. I've fixed this now locally
and will include the fix for 1.16.9, planned for tomorrow or the next
day.
But in any case gadfly is also buggy, as it's using the substvar
${source:Version} as the base for the binary version, which should
not include the binNMU version part. And using ${binary:Version}
which is what would appear to be correct is wrong as it will produce
recursive nested substitutions. The correct fix is to get the binary
version from dpkg-parsechangelog, prepend the epoch and pass that
to «dpkg-gencontrol -v», but that depends on a fixed dpkg 1.16.9.
thanks,
guillem
--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
