Re: GOP2-2b - Stable 2.16.x releases (dictator) (probable decision)

[John Mandereau]

Il giorno mer, 01/08/2012 alle 18.08 +0100, Graham Percival ha scritto:
> Hmm.  I can't answer this directly, but I'll pass along my
> considerations:
> 
> - if you try to compile GUB on debian unstable (or any other
>   recent distro), you will likely encounter odd compile failures.
>   These are important to fix at some point in time (otherwise we'd
>   be stuck GUB on ubuntu 10.04 only), and require a great deal of
>   knowledge of compilers and searching for solutions online.
>   OTOH, it might just work "out of the box" in which case it'll
>   just take 6-12 hours and then be working.

With my Fedora installation on that Core 2 Duo (x86_64), GUB Python
program has screwed up dependencies (it missed a lot of them) in a
consistent way across different Fedora versions, and when I reported it
I got not clue on debugging the intricated Python code of GUB, so I
think my next attempt at building GUB will be a chrooted Debian stable.
Now I've freed my main machine from lilypond-patchy-staging, I can more
easily build a GUB in background while doing other things, reporting how
it progresses  and asking for help if needed.


> - I think that supporting build numbers will be an easier
>   introduction to version number handling in GUB and our docs than
>   jumping straight into 4-tuples.  The first step is to make it
>   work in "make website", which is infinitely easier than trying
>   to do anything in GUB.  This is a relatively easy thing to fix,
>   so it might make sense to leave it for a relative beginner...
>   OTOH, the bug has existed for two years, so might as well tackle
>   it now.

I guess that if we want the release number in "make website", we want it
in HTML footers and manuals, don't we?


>   Also, David is quite likely to want to use build
>   numbers if they are available.
>   (whereas I'm happy to say "screw users" or "screw version
>   numbers" and either not bother updating with -2 if there's a
>   serious problem, or else bump to .x+1 one day after a .x
>   release; neither of those options are particularly ideal for a
>   stable branch)

IMHO patch level numbers are cheap, as it doesn't often happen to
release twice within two or three days.

Best,
John