Samuel Verschelde a écrit :
Le vendredi 28 octobre 2011 18:17:18, Anssi Hannula a écrit :
Hi!
Currently the updates policy [1] says that updates to a new version
should have 1.1.mga1 as release, and that cauldron should be bumped to
2.mga2 on those cases.
However, on the packages where version updates are done (e.g. wine,
flash-player-plugin, opera, firefox, chromium), doing that causes the
following chain of events on every update:
1. cauldron submission with 1.mga2
2. mga1 submission with 1.1.mga1
3. cauldron submission with 2.mga2
So the pkgs have to always be submitted to cauldron twice (unless one
skips right to 2.mga2, which would be rather confusing IMO).
I suggest we change the policy to say that package updates to stable
should use 1.mga1 release (i.e. no subrel), so that it goes:
1. cauldron submission with 1.mga2
2. mga1 submission with 1.mga1
and no extra submissions or commits required.
Then, if some extra fix is needed, one does:
1. cauldron submission with 2.mga2
2. mga1 submission with 1.1.mga1
WDYT?
[1] http://www.mageia.org/wiki/doku.php?id=updates_policy
I warned about this when we discussed about release 0 and subrels for updates,
but nobody reacted against at that time, so it ended in the written policy :)
See https://mageia.org/pipermail/mageia-dev/2011-July/007080.html
I don't see drawbacks to your change proposal, if no one does, it gets my
vote.
I agree as well. No sense in useless rebuilding.
Sometimes the version in cauldron is only an upstream bugfix to the
version in stable.
It that case, it is the only logical thing to do.
Since except for the versioning, the builds in cauldron and stable would
be identical.
That is the case for xkeyboard-config, where version 2.3 in cauldron is
an upstream bugfix for version 2.2 in mga1
(https://bugs.mageia.org/show_bug.cgi?id=2280)
I noticed that that was also essentially the case for a recent bug with
hunspell, even though upstream had bumped the major release number.
(https://bugs.mageia.org/show_bug.cgi?id=2272)
(Actually several upstream updates, one with an alternate release
major/minor designations. Essentially bugfix updates.)
So a lot of effort was made to provide an update by patches to fix a
bug, when just using the update in cauldron would have satisfied the
problem. With no risk of regression, since the requires were identical,
and many if not most of the dependant localisations were identical as well.
I'm not sure if the current policy, as written, prevents using the
version from cauldron as an update to stable, when the major/minor
release numbers differ.
Would be good to clarify / modify if necessary.
Best regards
Samuel Verschelde
Regards :-)
--
André
