kdecoration2 mess - std::function makes the world fun

[Scott Kitterman]

On Thursday, August 20, 2015 10:50:39 PM Maximiliano Curia wrote:
> ¡Hola Lisandro!
> 
> El 2015-08-20 a las 14:25 -0300, Lisandro Damián Nicanor Pérez Meyer 
escribió:
> >> After having spent quite some time today looking at the current
> >> kdecoration2 mess (tracked at
> >> https://bugs.kde.org/show_bug.cgi?id=350964 ,  at debian:796194,
> >> debian:796184 and probably many more)
> >> 
> >> The crash happens around when a std::function (a c++11 feature) is passed
> >> across a library boundary and some parts of it is built with gcc5 and
> >> others with 4.9.
> > 
> > Matthias: is there any chance to make this simbol part of the new ABI? In
> > this way stuff will not migrate to testing as it happened for us.
> 
> std::function seems to be a header only thing, so there is no related symbol
> in libstdc++6.
> 
> >> According to Matthias Klose, upstream actually gives no guarantees for
> >> for
> >> their abi for c++11, so I guess the steps forward should be:
> >> 
> >> 1) Immediately binNMU src:kdecoration2 in testing, then when that is
> >> done,
> >> binNMU src:kwin, src:oxygen and src:breeze all in testing. (that's for
> >> release team)
> > 
> > I don't know if we can binNMU something in testing with a testing chroot.
> > I
> > think it needs a sourcefull upload to t-p-u. RT, please confirm.
> 
> The release team asked us to do this after having a newer version in
> unstable blocked by a transition. So we need to request the binNMU after
> 2).
> >> 2) At the same time, do a library transition for libkdecoration2-5 and
> >> rename it to libkdecoration2-5v5 and binNMU the rdeps (the same as
> >> mentioned above)
> > 
> > I can do this if the RT ACKs it (all B-Ds are ready).
> 
> I'm preparing this upload.

FYI, I've just accepted it into Unstable, so hopefully it'll be there shortly.

Scott K