On ti, 2012-03-20 at 19:21 -0400, Rafaël Carré wrote: > Le 2012-03-20 17:53, Xavier Bachelot a écrit : > > On 03/20/2012 10:49 PM, Jean-Baptiste Kempf wrote: > >> On Tue, Mar 20, 2012 at 10:43:17PM +0100, Xavier Bachelot wrote : > >>> Was that expected ? > >> > >> Yes > >> http://git.videolan.org/?p=libbluray.git;a=commit;h=28d96f77ab8a774c8e2ef5b88a6ebced84794608 > >> > > > > What I meant is, isn't that an ABI breakage and shouldn't the soname be > > bumped to acknowledge that ? > > > > Regards, > > Xavier > > ABI didn't change, only API; so no problem.
That's how I see it too. Technically changing semantics of a function is ABI change. But semantics were not changed in this case. Application was never allowed to modify returned data. Adding const there just enforces already existing semantics at API level. This may cause compilation problems with C++ apps that store returned pointer to non-const pointer... But 0.2.2 is still backwards binary compatible with 0.2.1. - Petri _______________________________________________ libbluray-devel mailing list [email protected] http://mailman.videolan.org/listinfo/libbluray-devel
