Getting rid of wxwidgets2.6
Currently we have wxwidgets2.6 and wxwidgets2.8 in the archive. We've
had wxwidgets2.8 since lenny and it's now much more widely used than
2.6 in the archive. Upstream 2.6 no longer really gets any attention
(the last release was in 2007), while 2.8 had a point release about 6
So I believe it's time to migrate the remaining wxwidgets2.6-dependent
packages to wxwidgets2.8 and get rid of wxwidgets2.6. If you maintain a
package which depends on one of the wxwidgets2.6 packages, please at
least scan through the rest of this mail.
Further Reasons to do this now
It seems wxwidgets3.0 is approaching (an upstream dev says "beginning of
the next year" in a fairly recent blog post at
though they're also apparently still merging substantial new pieces of
code). We can't realistically support three major versions of wx at
once, and there are a number of newer applications which already need a
newer wx than 2.8, so there's going to be a strong demand for wx 3.0 to
be packaged when it finally gets released.
There are 6 Zack-months (that's about 8 in human months) to the proposed
freeze date for wheezy, so if we start now there's a realistic
possibility of having this done in time for wheezy.
Action so far
I migrated survex-aven (the wx-dependent package I co-maintain) to 2.8
before the squeeze release, so I'm at least practising what I preach!
I've been getting wxwidgets2.8 back into shape over the past few days -
there's an updated version of 2.8.10 in unstable with the RC bugs fixed
and low hanging fruit plucked - it's built successfully for all
supported architectures and is now just waiting to age enough to
I've also updated the packaging to the latest upstream release (2.8.12)
and picked the slightly more elevated fruit. I'll be uploading this to
experimental shortly - although it's nominally a minor version bump,
Torsten Landschoff noted in #586129 that:
However, there are also API changes at least in wxPython (some flags are
now passed differently for example) so I am not sure if one would want
to upload a new package right away.
(Torsten: Can you recall the details?)
This suggests caution is wise, as does the size of the diff between
2.8.10 and 2.8.12:
2029 files changed, 117347 insertions(+), 69829 deletions(-)
Quite a lot of that seems to be in generated files, but still...
I've gone through all the reverse dependencies of the 2.6 packages and
put the information on the wiki:
I'm going to start working through the list, but it's hard to test
applications I don't actually use, especially for the Python ones
where use of a removed or changed API won't get spotted until the
code path it is in is used. So please help out by testing any
applications you maintain or have a particular interest in. If
you are looking at a package, please just note your name and the
date in brackets after it to avoid duplication of effort.