On Nov 29, 2013, at 4:32 AM, Max Fest <[email protected]> wrote: > i got an issue with the lib, i get some random segmentation fault with my > program on basic operation. > During the call of a destructor for example. > > On the top of my backtrace i've got this : > > (gdb) frame 0 > #0 arena_dalloc_bin_locked (arena=arena@entry=0x7f4fc645b1c0, > chunk=chunk@entry=0x7f4de8000000, ptr=ptr@entry=0x7f4de817b800, > mapelm=<optimized out>) at > /home/server/dwcatacore3/dep/jemalloc/src/arena.c:1713 > 1713 } else if (run->nfree == 1 && run != bin->runcur) > (gdb) info locals > pageind = 379 > run = 0x7f4de8179000 > bin = 0x0 > bin_info = 0x1f551d0 > size = 1536 > > > So you can see that bin is NULL, and the call at bin->runcur will ran into a > segmentation fault. > > I use jemalloc 3.4.1 compiled with my project as a dependance. It run on a > Debian 7 x86_64. > > Is it a jemalloc related issue ? > The full backtrace : http://pastebin.com/w2KVAs8p
It’s likely that your application is corrupting page run state by double freeing an allocation. If you run a debug build of jemalloc and disable thread caching, jemalloc will probably fail an assertion when the double free occurs. Jason _______________________________________________ jemalloc-discuss mailing list [email protected] http://www.canonware.com/mailman/listinfo/jemalloc-discuss
