Re: Bug#759099: libcdio-dev: Causes every apt command to argue about unknown Multi-Arch type
- To: 759099@bugs.debian.org, dose-builddebcheck@packages.debian.org
- Subject: Re: Bug#759099: libcdio-dev: Causes every apt command to argue about unknown Multi-Arch type
- From: Axel Beckert <abe@debian.org>
- Date: Tue, 26 Aug 2014 21:23:13 +0200
- Message-id: <[🔎] 20140826192313.GN7356@sym.noone.org>
- In-reply-to: <CAEA6rAyRf-T4V=t4SWpCXHb22jbkH7ZTiuvLYnWVYs0HeCGbCA@mail.gmail.com>
- References: <87sikmdy65.fsf@c-cactus.deuxchevaux.org> <20140824095625.GA9661@jwilk.net> <CAEA6rAy3QreyJy7d_7+S8QxxCaHtJqPgbci4V-SFhvid5+j6eQ@mail.gmail.com> <20140824133144.GA321@debian.org> <20140824151614.GA29865@gaara.hadrons.org> <CAEA6rAyRf-T4V=t4SWpCXHb22jbkH7ZTiuvLYnWVYs0HeCGbCA@mail.gmail.com>
Control: clone -1 -2
Control: reassign -2 dose-builddebcheck 3.2.2-2
Control: retitle -2 The applications raised this exception : Format822.Type_error("Field Multi-Arch has a wrong value : no")
Hi,
as adsb mentioned on #debian-devel and #debian-buildd,
dose-builddebcheck also implements the wrong negative value
Julian Andres Klode wrote:
> >> > >> $ apt-cache policy foobar
> >> > >> W: Unknown Multi-Arch type 'no' for package 'libcdio-dev'
> >> > >> W: Unknown Multi-Arch type 'no' for package 'libiso9660-dev'
> >> > >> W: Unknown Multi-Arch type 'no' for package 'libudf-dev'
> >> > >
> >> > > According to deb-control(5), "no" is a valid value for the Multi-Arch
> >> > > field. (Although it's the default, so it would be probably better to omit
> >> > > the field.)
> >> >
> >> > The multi arch spec and APT use "none", not " no" - first time I heard of
> >> > "no". When was that changed, why, and why do I hear of it the first time
> >> > today?
> >>
> >> Clarification after checking source code: APT does not parse none
> >> either. But I still wonder why this was changed to "no" in dpkg when
> >> anyone else uses none (even APT internally uses none as the term, not
> >> "no").
> >
> > dpkg never changed the value, it's been “no” since the beginning, even
> > the MultiArch spec listed it as “no”, but it had some pretty recent
> > broken updates which changed that (those need to be reverted as I
> > mentioned in #732648), it was just not documented in deb-control(5),
> > which got fixed in dpkg commit c4ce860fdea16610e0c065cb265a24048a04c17c.
> >
> > See commit e215e2795f47feec839025972ac56e647ab04822 when this got
> > introduced in dpkg.
>
> OK thanks, it's always been called none outside of dpkg, though, as
> far as I can remember, the earliest source being from February 2011:
> https://lists.debian.org/debian-dpkg/2011/02/msg00051.html
>
> Maybe it was the case temporarily in some Ubuntu branches, I don't
> know the specifics.
For dose-builddebcheck the error can be reproduced on amd64 currently
like this:
$ dose-builddebcheck --deb-native-arch=amd64 /var/lib/apt/lists/ftp.ch.debian.org_debian_dists_experimental_main_binary-amd64_Packages /var/lib/apt/lists/ftp.ch.debian.org_debian_dists_sid_main_source_Sources
native-architecture: amd64
The applications raised this exception : Format822.Type_error("Field Multi-Arch has a wrong value : no")
This currently causes hanging experimental buildd queues on multiple
architectures. See e.g.
https://buildd.debian.org/status/package.php?p=libcdio&suite=experimental and
https://buildd.debian.org/status/package.php?p=zsh&suite=experimental
(Note to versions not being in sync, at least as of now.)
Regards, Axel
--
,''`. | Axel Beckert <abe@debian.org>, http://people.debian.org/~abe/
: :' : | Debian Developer, ftp.ch.debian.org Admin
`. `' | 1024D: F067 EA27 26B9 C3FC 1486 202E C09E 1D89 9593 0EDE
`- | 4096R: 2517 B724 C5F6 CA99 5329 6E61 2FF9 CD59 6126 16B5
Reply to: