After working for a few hours, I have glusterd wrecking itslef on a node. I already reported that for 3.4.0qa8, but at that time it happened only on an amd64 machine, while now I get it on an i386. Kernel trace shows no activity from glusterd. Attaching it with gdb raise a pthrerad related assertion failure in gdb. Obviously the glusterd memory is corrupted at that time. On 3.4.0qa8 I started investigating with electric-fence, and found an arg array to lack NULL termination. I did not have time to post a fix, but obviosuly a memory corruption byg remain in 3.4.0alpha. For now it seems it strikes only at mine, but who knows? It may bite someone lese differently. -- Emmanuel Dreyfus manu@xxxxxxxxxx
