Re: GC crash with 21.2.36 on Solaris 2.7

Ray> Could there be a GC bug in file_truename or some process I/O stuff? Ray> Lately, whenever I get a crash in GC, it seems to be always just after Ray> I've tried to do some kind of I/O like saving the file or auto-saving Ray> the file. How do I go about tracking this down? Since it's Ray> intermittent, I think doing a GC after every command (or whatever that Ray> option is) will take a long time. Yes, always_gc set to 1 slows down XEmacs enormously. But if you have a likely suspect, then that may be the best way to find a _reproducible_ crash, at which point it can be dissected, analysed and maybe even fixed. If you can't reproduce the crash, you never know whether you've fixed it. Here's how to do it without waiting an eternity. Build a debug xemacs, set a breakpoint at your suspect function (here probably Ffile_truename), run till you hit the breakpoint, then (gdb) set always_gc=1 (gdb) finish If you crash here reproducibly, congratulations! you've found it. If you return to gdb, better luck next time... Ben did some GC work on these sorts of functions, but I thought that was only 21.2. Martin