On Nov 5, 2011, at 12:47 PM, Mathieu Bouchard wrote: > Le 2011-11-05 à 00:50:00, Hans-Christoph Steiner a écrit : > >> The problem with changing t_widgetbehavior is that it breaks binary >> compatibililty, I think. > > Yes, and I already had made extra-crunchy hacks in GridFlow so that it can be > binary-compatible with both Vanilla and Extended while doing m_imp.h hacks. > This was because Extended introduced the quite unnecessary m_blobmethod field. > > So, I'd have to do another hack like that (but easier) to support L2Ork. I'm > not interested in distributing more binaries, so either Ivica distributes his > own GridFlow binaries, or preferably, I would have some code to auto-detect > L2Ork and act accordingly. > > Coïncidentally, both binary-compat problems have to do with (different parts > of) the comment-inlet hack used in the GFDP helpfile «template». > > There are similar binary-compat problems trying to deal with multiple > versions of Extended because GridFlow also links to GEM, which comes in > multiple binary-incompatible versions depending on the version of Extended. > > GF supports 3 different ABIs of GEM, but now there's a new ABI that it isn't > compatible with yet. It's the 3rd different GEM ABI to be labelled «0.93» so > far. (but I don't remember which versions of GEM are bundled with Extended).
Is that a Pd-extended issue or a Gem issue? I don't really know much about Gem's guts or C++ linking. I was under the impression this issue is related to C++ name mangling. .hc ---------------------------------------------------------------------------- "Making boring techno music is really easy with modern tools, but with live coding, boring techno is much harder." - Chris McCormick _______________________________________________ [email protected] mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
