On Tue, 2012-06-12 at 12:32 +0100, Daniel P. Berrange wrote: > On Tue, Jun 12, 2012 at 02:23:33PM +0300, Alon Levy wrote: > > Current bumped and age bumped. Current bumped twice since previous > > values (current = 1, age = 2) were illegal - age must be smaller or > > equal to current, since it's interpreted as meaning our library supports > > t interfaces [current, current - age]. > > > > > -m4_define([SPICE_CURRENT], [1]) > > +m4_define([SPICE_CURRENT], [3]) > > IIUC, this will cause an SONAME change, since SPICE_CURRENT is the part > used to form the binary name 'libspice-server.so.1'. So this is an > ABI break requiring all apps to recompile again :-( > Yes, it changes the SONAME. I think that's exactly right - there are added interfaces, so anyone compiling against the new version should require it specifically, hence it should link against NEW-SONAME and not OLD-SONAME. But I thought it doesn't break existing users. If it does that's not what I meant. I thought the important name was that before the .so, i.e. libspice-server, and that the dynamic linker will look for .1 and then if none found will use anything newer. > IIUC, SPICE is supposed to be ABI stable, which means we should never > be changing the SONAME for any release. Then libtool's info page is garbage. I guess I should readup on what is the dynamic linker looking at. > > > m4_define([SPICE_REVISION], [0]) > > -m4_define([SPICE_AGE], [2]) > > +m4_define([SPICE_AGE], [3]) > > > > AC_INIT(spice, [SPICE_MAJOR.SPICE_MINOR.SPICE_MICRO], [], spice) > > > Regards, > Daniel _______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/spice-devel