Hi Sean,
On Tue, Mar 28, 2006 at 05:22:51AM -0500, sean finney wrote:
> i got a grave bug reported against nagios recently (#359637), because
> nagios is no longer installable in unstable. it looks like
> someone's bin-NMU'd it on 2006/03/20, along with the other
> arch:any packages.
> my best guess is that this has to do with the recent mysql
> versioning vs our versioning of symbols fiasco with libmysqlclient15,
> but there are a couple questions that have me a bit concerned, or
> at least curious:
> - am i correct?
> - why was no bug reported against my package before the binNMU?
> - why was i not informed of the binNMU afterwards?
Because ideally, such binNMUs should be completely transparent to
maintainers. Except when something breaks, like this. :)
> the big problem here is that the new packages have a bumped
> debian revision. of course this makes sense (and a future
> upload by me will make these packages disappear), but it
> has in the meantime silently rendered nagios uninstallable
> because there is a "sourceful" (sorry, not quite sure what the
> correct terminology is) depends on nagios-common. that is:
> nagios-mysql depends on nagios-common (= ${Source-Version})
> so when nagios-mysql is bumped without bumping nagios-common, we
> have our current situation.
Which is no worse than the previous situation, of nagios-mysql being
uninstallable because libmysqlclient15 is no longer available.
> my guess is that there are other packages which may have a similar
> problem.
Yes, arch: any packages with strict dependencies on arch: all packages are a
common problem. We need a systemic solution for this to be implemented in
dpkg.
On Tue, Mar 28, 2006 at 08:13:51PM +0200, Andreas Metzler wrote:
> At least I as a package maintainer am not very interested in:
> bug#465656 mysql breakage.
> dear maintainer please do not make a sourceful upload, we (d-r)
> willhandle this using binNMUs.
> followed by a
> bug#465656 fixed, we made the binNMUed
> two hours later.
And I'm not interested in having to close such bugs. :)
> Imho being incompatible with binNMUs is a bug, the dependency clearly
> is too strict, as nagios-mysql 2:1.3-cvs.20050402-12+b1 surely _would_
> work with nagios-common 2:1.3-cvs.20050402-12. Wouldn't
> Depends: nagios-common (=> ${upstreamversion}), nagios-common (<= ${upstreamversion}.0)
> work for you?
In the general case one cannot know, a priori, that a given version of a
package will be compatible with future Debian revisions of a related
package. I don't think maintainers that are concerned about such
incompatibilities should be asked to change their package dependencies just
to support binNMUs. The right solution here *is* to enhance dpkg-dev so
that it knows the difference between source and binary versions.
--
Steve Langasek Give me a lever long enough and a Free OS
Debian Developer to set it on, and I can move the world.
vorlon@debian.org http://www.debian.org/
Attachment:
signature.asc
Description: Digital signature