Re: version number when packaging a new upstream release
On Thu, 06 Oct 2016, Adrian Bunk wrote:
> I gave the a rationale in the following paragraph:
> In the general case it is even possible that the package was removed
> from unstable, but later someone ITPs 6:0.8.18-1 into unstable. At that
> point the version in oldstable would be higher than the version in
This is a corner-case that should be handled as such. There's no reason
for us to ignore the availability of a newer version in newer releases.
I claim that we don't have to make complicated rules when the situation is
simple because all newer releases have newer upstream versions.
> Another rationale I gave in the email you were replying to:
> Precedents in DSAs also suggests -0+deb7u1
Well, they had to make a choice and they picked something that respects
the rules we have. If we want to shoot for consistency (it's certainly a
good idea), then we have to document the result of this discussion in
the developers reference.
> It is beneficial when you can see from the version number what a package is:
> -1 - original -1 package, not modified by any updates to stable
> -0.1 - original -0.1 package, not modified by any updates to stable
> -1+deb7u1 - first update to the -1 package in Debian 7
> -0.1+deb7u1 - first update to the -0.1 package in Debian 7
> -1~deb7u1 - backport of the -1 package to Debian 7
> -0.1~deb7u1 - backport of the -0.1 package to Debian 7
> -0+deb7u1 - none of the above, not based directly on a version of
> the package that was ever in unstable
I don't think any of this is reliable enough to be useful. A -1 package
does not need to have started its life in unstable... the fact that a new
upstream version was NMUed does not bring any useful information.
And what's the technical difference between a version packaged for
unstable and then backported for oldstable and a version packaged directly
for oldstable? None. So why do we want to use different version numbers?
If we want anything reliable about the package's history, we have to read
In any case, I agree that simple consistent rules are better and thus
we should pick either -1~deb7u1 or -0+deb7u1.
> Do you have any rationale why you think -1~deb7u1 would be better
> than -0+deb7u1?
My preference goes for the former because it matches the logic of
backported packages and thus does not introduce a new concept while
-0+deb7u1 is not something we use in another context.
Raphaël Hertzog ◈ Debian Developer
Support Debian LTS: http://www.freexian.com/services/debian-lts.html
Learn to master Debian: http://debian-handbook.info/get/