[Debian-zh-dev] Bug#975045: Bug#975045: galternatives fails to build for binnmu

Sven Mueller sven.mueller72 at gmail.com
Wed Nov 18 17:28:29 GMT 2020

We are actually rebuilding all Debian packages and we do tag them as binary
nmu if we rebuild then again (in theory, this only happens in specific
cases, but sometimes the automation triggered too often). But in theory at
least: yes, there are reasons to do binary NMUs for arch:all packages as

(We also would tag the version numbers similarly if we have a local patch,
but IIRC, that wasn't the case here)

Yangfl <mmyangfl at gmail.com> schrieb am Mi., 18. Nov. 2020, 16:42:

> Sven Mueller <sven.mueller72 at gmail.com> 于2020年11月18日周三 下午6:39写道:
> >
> > Package: galternatives
> > Version: 1.0.7
> > Tags: patch
> >
> > debian/rules compares the version from debian/changelog with the version
> reported by the INFO dict. However, this fails if a binnmu is at play:
> >
> > A binnmu adds a changelog entry with +bX (X being an integer), but
> generally speaking, rebuilds might add almost any +<whatever> suffixes.
> >
> > I suggest to change debian/rules as follows to allow for binnmus:
> >
> > Add:
> >
> > BASE_VERSION=$(shell dpkg-parsechangelog \
> >       sed -rne 's/^Version: ([0-9.]+)(\+.*)?/\1/p')
> >
> > And compare to that instead of DEB_VERSION_UPSTREAM
> >
> > This is similar to what is done for DEB_VERSION_UPSTREAM in pkg-info.mk,
> but acknowledges the fact that galternatives is a native Debian package and
> thus the whole version from the changelog is recognized as the upstream
> version.
> >
> > For 1.2-1+b1, DEB_VERSION_UPSTREAM would have 1.2, as it strips the last
> - and everything after it.
> >
> > Patch attached, this time I'll refrain from submitting anything to
> salsa... I caused too much chaos with the previous commits already (I hope
> I cleaned it up though).
> Wait, does arch: all package ever has a chance to do binnmu?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://alioth-lists.debian.net/pipermail/chinese-developers/attachments/20201118/5eddb48b/attachment.html>

More information about the Chinese-developers mailing list