Bug#862178: Bug#859912: Bug#862178: problem with packages build-depending on locales in experimental
On Thu, 11 May 2017 at 15:33:41 +0200, Daniel Baumann wrote: > On 05/11/17 01:27, Aurelien Jarno wrote: > > The build daemons use aspcud as resolver for the experimental > > distribution, as aptitude is not able to handle many cases. > > why not using apt? Because the expected behaviour for experimental and *-backports is not the same as all the other suites. In unstable, testing, stable, (old*)stable and security suites, it is sufficient to enable some apt sources, and take the newest version of each package available from any source. However, that's a bad heuristic for backports. For example, libglib2.0-0 has a jessie backport, presumably because some package that was backported needs it. However, GLib users in backports don't necessarily need that version: for example, libostree and flatpak can work with the GLib from jessie. Preferring to build against the version in jessie, and only using the version from backports if a versioned build-dep makes it strictly necessary, means that you don't need to upgrade GLib (potentially destabilizing unrelated packages) just to get the flatpak backport. Similarly, in experimental, just because a build-dependency is available in experimental at a newer version, that doesn't necessarily mean that all other experimental packages that depend on that library need the newer version. S
Bug#862178: Bug#859912: Bug#862178: problem with packages build-depending on locales in experimental
On 05/11/17 01:27, Aurelien Jarno wrote: > The build daemons use aspcud as resolver for the experimental > distribution, as aptitude is not able to handle many cases. why not using apt? Regards, Daniel
Bug#862178: Bug#859912: Bug#862178: problem with packages build-depending on locales in experimental
On Thu, May 11, 2017 at 08:08:06AM +0900, Mike Hommey wrote: > [...] I'm all for open-infrastructure-locales-c.utf-8 being removed > from the archive because it's a broken package that does nothing > useful [...] I just had pygobject broken in experimental by this bug too. Any objection to me (or someone else if they would like to) filing an RoQA RM bug? Cheers, -- Iain Lane [ i...@orangesquash.org.uk ] Debian Developer [ la...@debian.org ] Ubuntu Developer [ la...@ubuntu.com ] signature.asc Description: PGP signature
Bug#859912: Bug#862178: problem with packages build-depending on locales in experimental
On Thu, May 11, 2017 at 06:13:13AM +0300, Adrian Bunk wrote: > On Thu, May 11, 2017 at 08:08:06AM +0900, Mike Hommey wrote: > > On Wed, May 10, 2017 at 04:44:07PM +0300, Adrian Bunk wrote: > > > On Tue, May 09, 2017 at 04:08:04PM +0200, Daniel Baumann wrote: > >... > > > > For the rational why o-i-locales-c.utf-8 is usefull, see its manpage: > > > > > > > > https://manpages.debian.org/unstable/open-infrastructure-locales-c.utf-8/locales-c.utf-8.7.en.html#Use_Case > > > >... > > > > > > The rationale in the manpage is nonsense. > > > > To me, it's all a long text just trying to say without really saying it > > that the sole purpose is to reduce disk space. IOW, this could be > > fulfilled by use of dpkg's path-exclude. > >... > > It does not even reduce disk space. > > The C.UTF-8 locale is provided by the essential libc-bin package. It does save the disk space from the locales/locales-all packages. But dpkg's path-exclude would have the same effect. Mike
Bug#859912: Bug#862178: problem with packages build-depending on locales in experimental
On Thu, May 11, 2017 at 08:08:06AM +0900, Mike Hommey wrote: > On Wed, May 10, 2017 at 04:44:07PM +0300, Adrian Bunk wrote: > > On Tue, May 09, 2017 at 04:08:04PM +0200, Daniel Baumann wrote: >... > > > For the rational why o-i-locales-c.utf-8 is usefull, see its manpage: > > > > > > https://manpages.debian.org/unstable/open-infrastructure-locales-c.utf-8/locales-c.utf-8.7.en.html#Use_Case > > >... > > > > The rationale in the manpage is nonsense. > > To me, it's all a long text just trying to say without really saying it > that the sole purpose is to reduce disk space. IOW, this could be > fulfilled by use of dpkg's path-exclude. >... It does not even reduce disk space. The C.UTF-8 locale is provided by the essential libc-bin package. > Mike cu Adrian -- "Is there not promise of rain?" Ling Tan asked suddenly out of the darkness. There had been need of rain for many days. "Only a promise," Lao Er said. Pearl S. Buck - Dragon Seed
Bug#859912: Bug#862178: Bug#859912: Bug#862178: problem with packages build-depending on locales in experimental
On 2017-05-11 08:08, Mike Hommey wrote: > On Wed, May 10, 2017 at 04:44:07PM +0300, Adrian Bunk wrote: > > On Tue, May 09, 2017 at 04:08:04PM +0200, Daniel Baumann wrote: > > >... > > > Can anybody have a look at this why it's failing? Bug with references is > > > #859912 > > >... > > > > Your package shouldn't exist, this is the main problem here. > > I won't say I disagree with this, but... > > > And the potential breakage is not limited to buildds. > > > > Package: open-infrastructure-locales-c.utf-8 > > Provides: locales, locales-all, locales-c.utf-8 > > > > Your package claims to provide locales-all without actually providing it. > > This is a bug in your package. > > > > If I have open-infrastructure-locales-c.utf-8 already installed and then > > install something that depends on locales-all, your broken package would > > fulfill the dependency causing locales-all not getting installed. > > See https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=859912#48 > > Disregarding the existence of open-infrastructure-locales-c.utf-8, we > do have a problem if people can't reproduce buildds failures outside of > buildds. > > I'm not a sbuild user, but pbuilder doesn't end up installing that > package. According to Daniel's message, it's not happening with sbuild > either. So what the hell? The build daemons use aspcud as resolver for the experimental distribution, as aptitude is not able to handle many cases. I guess you can reproduce the issue by passing --build-dep-resolver=aspcud to sbuild and by adding the following line to sbuild.conf: $aspcud_criteria = '-removed,-changed,-new,-count(solution,APT-Release:=/experimental/)'; -- Aurelien Jarno GPG: 4096R/1DDD8C9B aurel...@aurel32.net http://www.aurel32.net
Bug#859912: Bug#862178: problem with packages build-depending on locales in experimental
On Wed, May 10, 2017 at 04:44:07PM +0300, Adrian Bunk wrote: > On Tue, May 09, 2017 at 04:08:04PM +0200, Daniel Baumann wrote: > >... > > Can anybody have a look at this why it's failing? Bug with references is > > #859912 > >... > > Your package shouldn't exist, this is the main problem here. I won't say I disagree with this, but... > And the potential breakage is not limited to buildds. > > Package: open-infrastructure-locales-c.utf-8 > Provides: locales, locales-all, locales-c.utf-8 > > Your package claims to provide locales-all without actually providing it. > This is a bug in your package. > > If I have open-infrastructure-locales-c.utf-8 already installed and then > install something that depends on locales-all, your broken package would > fulfill the dependency causing locales-all not getting installed. See https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=859912#48 Disregarding the existence of open-infrastructure-locales-c.utf-8, we do have a problem if people can't reproduce buildds failures outside of buildds. I'm not a sbuild user, but pbuilder doesn't end up installing that package. According to Daniel's message, it's not happening with sbuild either. So what the hell? > >... > > For the rational why o-i-locales-c.utf-8 is usefull, see its manpage: > > > > https://manpages.debian.org/unstable/open-infrastructure-locales-c.utf-8/locales-c.utf-8.7.en.html#Use_Case > >... > > The rationale in the manpage is nonsense. To me, it's all a long text just trying to say without really saying it that the sole purpose is to reduce disk space. IOW, this could be fulfilled by use of dpkg's path-exclude. https://raphaelhertzog.com/2010/11/15/save-disk-space-by-excluding-useless-files-with-dpkg/ I'm all for open-infrastructure-locales-c.utf-8 being removed from the archive because it's a broken package that does nothing useful, but I still think we have a problem at hand when we can't reproduce what buildds are doing (and they're doing something weird). Mike
Bug#859912: Bug#862178: problem with packages build-depending on locales in experimental
On Tue, May 09, 2017 at 04:08:04PM +0200, Daniel Baumann wrote: >... > Can anybody have a look at this why it's failing? Bug with references is > #859912 >... Your package shouldn't exist, this is the main problem here. And the potential breakage is not limited to buildds. Package: open-infrastructure-locales-c.utf-8 Provides: locales, locales-all, locales-c.utf-8 Your package claims to provide locales-all without actually providing it. This is a bug in your package. If I have open-infrastructure-locales-c.utf-8 already installed and then install something that depends on locales-all, your broken package would fulfill the dependency causing locales-all not getting installed. >... > For the rational why o-i-locales-c.utf-8 is usefull, see its manpage: > > https://manpages.debian.org/unstable/open-infrastructure-locales-c.utf-8/locales-c.utf-8.7.en.html#Use_Case >... The rationale in the manpage is nonsense. As the manpage already correctly states: The C.UTF-8 locale is included in the libc-bin package which is a package marked essential and thus always present on any Debian based system. In other words: C.UTF-8 can always be used on a Debian system, and any package depending on locales-all can be expected to require locales other than C.UTF-8. Debian based systems expect a UTF-8 capable locale to be used. No, they do not. Even the default locale is not UTF-8. On minimal systems such as servers and containers system administrators often prefer to use the C.UTF-8 locale. If a system administrator prefers to use the C.UTF-8 locale, he can already use it without installing any non-essential package. > Regards, > Daniel cu Adrian -- "Is there not promise of rain?" Ling Tan asked suddenly out of the darkness. There had been need of rain for many days. "Only a promise," Lao Er said. Pearl S. Buck - Dragon Seed
