Also sending the message to the bug report. ----- Forwarded message from Sebastian Ramacher <sramac...@debian.org> -----
Date: Fri, 16 May 2014 15:52:04 +0200
From: Sebastian Ramacher <sramac...@debian.org>
To: Dmitry Smirnov <only...@debian.org>
Subject: Re: Bug#748304: tupi: not binNMU safe
Message-ID: <20140516135204.gb6...@ramacher.at>
On 2014-05-16 23:09:31, Dmitry Smirnov wrote:
> On Fri, 16 May 2014 00:29:20 Sebastian Ramacher wrote:
> > tupi is not binNMU safe. After the rebuild for the libav 10 transition,
> > the package is no longer installable:
> >
> > [...]
> >
> > The following packages have unmet dependencies:
> > tupi : Depends: tupi-data (= 0.2+git04-1+b1) but 0.2+git04-1 is to be
> > installed E: Unable to correct problems, you have held broken packages.
>
> Sorry, Sebastian, but I don't understand what is the problem here.
> Like many other presumably NMU-safe packages, tupi/any Depends on
> tupi-data/all (= ${source:Version}). What can I possibly fix here?
>
> Now after binNMU tupi somehow Depends on "tupi-data (= 0.2+git04-1+b1)".
> How could this happen?
>
> I understand this would be the case for arch-any package depending on
> "arch-all (= ${Source-Version})" but there is a lintian warning for this so I
> would hardly miss it and it is not the case anyway.
tupi-data (= ${source:Version}) in tupi's Depends is not the problem
here. This is fine. The problem is that tupi-data (= ${binary:Version})
appears in tupi's ${misc:Depends}.
When built for an usual upload, ${misc:Depends} looks like
$ head -1 debian/tupi.substvars
misc:Depends=tupi-data (= 0.2+git04-1)
which is okay, but if binNMUed it becomes
$ head -1 debian/tupi.substvars
misc:Depends=tupi-data (= 0.2+git04-1+b1)
which is wrong.
This is due to the dh_installdocs override. Note that if you remove the
override, you need to handle symlink to directory conversions.
Cheers
--
Sebastian Ramacher
----- End forwarded message -----
--
Sebastian Ramacher
signature.asc
Description: Digital signature
