As of current HEAD, the GC version of gnash triggers NO memory corruptions in our testsuite. Still there are two failures under misc-ming.all:
FAIL: place_and_remove_object_testrunner: _root.sh1 != undefined [place_and_remove_object_test.c:76] FAIL: place_and_remove_object_testrunner: expected: undefined obtained: _level0 [place_and_remove_object_test.c:77] I'll track them out, anyway the point is that our testsuite seems unable to find more memory corruption, so it's time to try it against other movies, your offline tests for example. Finding memory corruptions is the best way to make the GC version more stable. Basically, if gnash segfaults (and doesn't with the RC version) the procedure I use to find out what's being NOT properly marked is running valgrind, which gives you useful information about illegal memory accesses, in particular which piece of code allocated what is being referenced now. That place of code is usually a class method, with the class failing to properly mark that resource as reachable. Reports welcome :) --strk; () ASCII Ribbon Campaign /\ Keep it simple! _______________________________________________ Gnash-dev mailing list [email protected] http://lists.gnu.org/mailman/listinfo/gnash-dev
