Em 05/01/2010, às 02:05, Matt Rogers escreveu: > On Saturday 02 January 2010 07:31:45 pm Pau Garcia i Quiles wrote: >> On Sat, Jan 2, 2010 at 9:02 PM, Roman Jarosz <kedge...@gmail.com> wrote: >>> On Sat, 02 Jan 2010 17:05:28 +0100, Pau Garcia i Quiles >>> >>> <pgqui...@elpauer.org> wrote: >>>> On Sat, Jan 2, 2010 at 3:28 PM, Roman Jarosz <kedge...@gmail.com> wrote: >>>>> Hi, >>>>> >>>>> Boiko, here is the first pile of libmsn fixes. >>>>> >>>>> The 4th patch was not included intentionally. >>>> >>>> Regarding patch >>>> 0005-Remove-ifdef-LIBMSN_INBOX_URL_ENABLED-and-make-pure-.patch, I >>>> think you do not need to bump soname because the method is virtual. >>>> See http://tinyurl.com/ydfphzy >>> >>> Hm, I don't see anything about pure virtual functions >> >> From http://tinyurl.com/ydfphzy : >> >> "You cannot... >> [...] >> For virtual member functions: >> * add a virtual function to a class that doesn't have any virtual >> functions or virtual bases. >> * add new virtual functions to non-leaf classes as this will break >> subclasses. See below for some workarounds or ask on mailing lists. >> [...]" >> >> In this case, that's a leaf class and it already has virtual methods >> (i. e. there is a virtual table already), so the change it's >> binary-compatible. >> >>> , but the bump is >>> >>> also because >>> the older version (0.2.0) could have LIBMSN_INBOX_URL_ENABLED on/off so >>> this >>> will force Kopete recompilation and will fix all the unexpected crashes. >> >> Soversion was changed from 0.1.0 to 0.2.0 because >> LIBMSN_INBOX_URL_ENABLED broke binary compatibility even when >> LIBMSN_INBOX_URL_ENABLED was OFF, due to the addition of data members >> (search the archives of this mailing list, I explained this in great >> detail a few weeks ago). >> >> Bump soname just to force Kopete to be rebuilt is a very bad practice >> and I oppose. >> >> Bumping soname without a need makes my life as the Debian maintainer >> of libmsn more complicated because I need to submit the package and go >> through the FTP-masters queue again, which may take several weeks. >> Please don't. > > Sorry, but we're not responsible for Debian project process issues. We *are* > responsible for making sure we don't have stupid bugs and crashes in our > code. > If it takes a long time to get your stuff uploaded into Debian, that's not > our > problem. > > I'm in favor of an soversion bump if it makes our job as library users > easier, > then I'm all for it.
Totally agree. I'm going to apply the soname bump (I just hope Salem won't kill me when he is back ;) Cheers Boiko _______________________________________________ kopete-devel mailing list kopete-devel@kde.org https://mail.kde.org/mailman/listinfo/kopete-devel
