Hi, I'm probably the LilyPond developer most involved with GUILE 2.0 migration and I'm pretty annoyed at the current situation and the manner GUILE developers deal with it.
Several months back even Richard Stallman intervened and stressed the importance of getting LilyPond moved to GUILE 2.0. Like several times before, GUILE developers promised to get actively involved only to drop out of the discussion once they were provided with instructions, an up-to-date branch/source to work with and current problem descriptions. The current situation is such that 2.0 garbage collection API is unreliable (see GUILE bug report <URL:http://debbugs.gnu.org/cgi/bugreport.cgi?bug=19883> with the basic recommendation "don't try using the smob mark mechanism any more" but no real resolution). It may well be that the current workarounds implemented in LilyPond may be successful. However, this is hard to test since there is _no_ released version of GUILE 2.0 where the encoding problems in issues <URL:http://debbugs.gnu.org/cgi/bugreport.cgi?bug=20200> (workaround in LilyPond codebase, will get fixed in 2.0.12) and <URL:http://debbugs.gnu.org/cgi/bugreport.cgi?bug=20209> (workaround in LilyPond codebase, will get fixed in 2.0.12) and <URL:http://debbugs.gnu.org/cgi/bugreport.cgi?bug=20302> (unfixed so far, and since this usage _was_ already a workaround for previous problems and reverting back to the old code does not work either, this remains a roadblock) have been addressed. With the ongoing trail of suggested workarounds failing for new reasons, there has not been the possibility to actually get to the stage where it would be possible to do any serious testing with GUILE 2.0, like running the regtest suite. So it is very likely that there are more surprises lurking (particularly regarding garbage collection) once the GUILE developers get around to fixing the pending bugs in the bytevector stream port implementation. Or get around to actually following on their promises and try working on figuring out why the old workarounds for getting GUILE reproduce a byte stream stopped working. But since they decided to break them anyway in GUILE 2.1 it would likely make more sense to make the "please use bytevectors for this from now on" approach actually work and then let LilyPond switch to this mechanism once there is a working version of GUILE released with them. By the way: I was of the impression that TeXmacs did not work with GUILE 2.0 either. Has this changed? Is it also going to get removed from Debian? -- David Kastrup -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
