On 2007/07/06 15:49, Matthias Kilian wrote: > Sending some discussion with Kurt back to ports@ (no bounce, to avoid > confusion)... > > >> And for boehm-gc itself, libgc really needs a major version bump > >> (to 2.0), IMHO. > > Yes! Otherwise existing programs will stop working (due to > > the missing -pthread needed now in the exe's). > > Apart from the -pthread thing, the boehm-gc update seems far from being > complete: > > - amd64 support: yet unclear > - sparc64 support: no, according to Stuart > > I don't know wether boehm-gc 6.2 works on those archs, though.
w3m is marked broken there, but I suspect it may be the boehm-gc port that has the real problem. (iirc, segfaults in w3m on all but the simplest pages. hmm, wouldn't surprise me if it segfaults whenever GC actually _does_ collect any garbage). > +share/gc/porting.html I'm far from knowing enough about memory layout to work through this, but I suspect it has quite a few of the answers (though it doesn't go into threads very much). > Kurt asked me to provide a -uN type cvs diff (at end of this mail, > with the library bump included and gc.man @commented out again). (minor nit, but patch-doc_gc_man has a date that has changed) > He also noticed that the port doesn't honor CFLAGS and that amd64 > regress fails: > > > /usr/obj/ports/boehm-gc-7.0/gc-7.0/.libs/gctest gets stuck > > in a tight loop on amd64 in: > > 0x000000004af949d3 in GC_mark_from () from > /usr/obj/ports/boehm-gc-7.0/gc-7.0/.libs/libgc.so.1.2 Seen here too. > NOT_FOR_ARCHS= hppa powerpc vax Has anybody got anything working reliably with the existing 6.2 port on !i386? If not then updating to 7.0 with ONLY_FOR_ARCHS, ugly though it is, might satisfy those wanting Inkscape without making things much worse than they are, then it can be worked on properly, and gives us another app to test it with...
