On Thu, 18 Dec 2014, Zbigniew Jędrzejewski-Szmek wrote:
The plan is to make wxPython 3.0 co-installable with wxPython 2.8, similarly to what has been done with the wxWidgets packages (wxGTK and wxGTK3).
What are the issues involved in updating everybody to wxPython 3.0?
It is certainly doable. Debian did it recently in Jessie, so we can probably look there to see which packages needed patching. The main difference is that they are still using the GTK2 implementation. Upstream wxPython only recently enabled the GTK3 implementation, so I'm a little worried that we may come across some bugs.
Are you thinking that we should just go for it and upgrade everybody?
I really don't know. That's why I was asking.
Making them parallel installable is the kind of work which generates more work later on. I think it might be worthwhile to put the new version in a buildroot override and rebuild the dependent packages and see what falls out. It might be less work to fix the few problematic packages.
Well, none of the dependent packages should need to be rebuilt. Instead, we would need to build the new wxPython and then test each dependent package to see if anything breaks. That is certainly doable though.
That's even better. repoquery tells me that there are 35 packages requiring wxPython. This seems like a managable number to test, even for one or two people.
Okay, that's what I'll plan to do then, unless there are major objections. I'll probably need a bit of help from the package maintainers, though, as I'm sure I won't be able to figure out how to fully test all of the dependent packages.
Thanks, Scott