On 03/25/2014 04:12 AM, Jürgen Spitzmüller wrote:
2014-03-25 9:02 GMT+01:00 Vincent van Ravesteijn:
I wonder wether it is a good idea to backport things to 2.0.8
_before_ it is in master. You can be almost sure that noone has
tested or even tried or maybe even looked at the patch if it is
only in 2.2-staging (except the branch maintainer maybe).
I suppose it depends on the patch (and the severity of the problem). I
agree that we need to be extra careful with commits to branch,
especially to the (planned) concluding release of a cycle. But then,
the normal situation (not immediately before a release, as now, but in
normal development cycle) is just so: If patches are in master, they
won't get tested, since nobody uses master for real work. They only
get tested when they are in branch. So I'd say the situation is
(normally) just reversed to what you say.
This is why I started to use the stable branch for real work when I
was stable release manager. Now, I have switched to 2.1.x (some time
ago, as a matter of fact), so this bug at least escaped _me_. But
then, it was detected, after all.
Yes, this seems to me to be a fairly normal case of something having
been committed to stable branch, then not working as intended, so the
bug is detected and then, hopefully, fixed. I'm still using 2.0.x for my
work, for this very reason---though I haven't had to reconfigure
recently. It's also normal that, late in the cycle, the divergences
between stable and devel cause unexpected problems.
The problem itself has been fixed. We just needed to remove the
--binary-dir bit.
Richard
