On Mon, Jul 19, 2010 at 9:10 AM, David Kirkby <david.kir...@onetel.net> wrote: > Yes - see below. > > sage: quit > Exiting Sage (CPU time 0m0.29s, Wall time 0m3.52s). > > Program received signal SIGSEGV, Segmentation fault. > 0xffffffff7e6627ec in _free_unlocked () from /lib/64/libc.so.1 > (gdb) bt > #0 0xffffffff7e6627ec in _free_unlocked () from /lib/64/libc.so.1 > #1 0xffffffff7e662784 in free () from /lib/64/libc.so.1 > #2 0xffffffff78c13154 in ?? () > #3 0xffffffff78c13154 in ?? () > Backtrace stopped: previous frame identical to this frame (corrupt stack?) > (gdb) > > I don't know how to approach this.
Well, under Linux I would try valgrind next; but evidently valgrind doesn't work under Solaris. I'm guessing the problem is some sort of malloc heap corruption. If so, here are some tips on debugging that. Is there a way under Solaris to turn on some sort of memory debugging in libc? (For example, glibc does some extra heap corruption checks if the MALLOC_CHECK environment variable is set.) Otherwise, here's a list of malloc debug libraries. Probably some of them work on Solaris. http://www.fortran-2000.com/ArnaudRecipes/FreeMemoryDB.html If that fails, here's a technique I've used for pinpointing crash locations. It works if the crash is sufficiently reproducible (happens the same way on every run). Run "sage -gdb". Once you get the prompt, break into the debugger. Set breakpoints on a few frequently-called functions (they don't have to have anything to do with your problem, "frequently-called" is the real requirement). Type "continue 10000". Now you're back at the Sage prompt; exit Sage. If you hit a breakpoint, type "continue 10000" again. Repeat until the crash. Now you know that the crash happens, say, between 20,000 and 30,000 breakpoints after you exit Sage. Now you can restart Sage, set the same breakpoints, and do binary search to find, eventually, that the crash happens between breakpoint 23,456 and 23,457. Now restart once again, do "continue 23456", and then start stepping through the code line-by-line -- you will be fairly close to the crash, so hopefully this won't take too long. Carl -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
