On Sunday, August 16, 2015 05:03:04 PM Scott Kitterman wrote: > On Thursday, August 13, 2015 09:39:43 AM Stéphane Glondu wrote: > > Control: tags -1 + confirmed > > Control: severity -1 serious > > > > On Fri, 03 Jul 2015 13:14:20 +0000 Matthias Klose <d...@debian.org> wrote: > > > - Decide if the symbols matching __cxx11 or B5cxx11 are part of the > > > > > > library API, and are used by the reverse dependencies of the > > > library. > > > > There is at least one such symbol: _ZN5Smoke8classMapB5cxx11E > > > > > - If there are no reverse dependencies, it should be the package > > > > > > maintainers decision if a transition is needed. However this might > > > break software which is not in the Debian archive, and built > > > against these packages. > > > > There are many reverse dependencies. > > > > > - If a library transition is needed, please prepare for the change. > > > > > > Rename the library package, append "v5" to the name of the package > > > (e.g. libfoo2 -> libfoo2v5). Such a change can be avoided, if you > > > have a soversion bump and you upload this version instead of the > > > renamed package. Prepare a patch and attach it to this issue (mark > > > this issue with patch), so that it is possible to NMU such a > > > package. [...] > > > > TODO > > I looked and the non-optional symbol that goes missing is: > > c++filt _ZN5Smoke8classMapE > Smoke::classMap > > This is only used in qt4-perl according to codesearch.d.n. I've locally > updated smokegen and verified qt4-perl will still build against it. Since > only one rdepend is actually affected by this transition, I think it'd be > easier to just upload smokegen with the fixed symbols file, but no v5 > package bump and then binNMU qt4-perl. There's enough else going on that > bending the normal way of doing things makes sense in this case (at least > to me). > > Scott K
I see Ubuntu did a v5 and that's really the right way to do it, so I guess I'll do that. Working on it. Scott K
