Index

Show enters and exits. Hide enters and exits.

09:57:32Maledictushow to label a ticket on github?
11:41:52rueMaledictus: Think you need to be a project member for that.
11:59:13dbussinkMaledictus: i see more similar reports that libexecinfo is crashing
11:59:34dbussinkMaledictus: http://varnish-cache.org/ticket/548 for example
14:43:10Maledictusdbussink: thanks. will look into that :)
14:43:21Maledictusrue: ok
16:13:16Maledictusyay, I found the culprit. updated the ticket
16:14:45MaledictusAnd I have another nice backtrace. It's not unmangled it seems: http://bsdpaste.bsdgroup.de/13886
17:05:31dbussinkMalediktus: what was the problem?
17:11:17Malediktusseems like gcc on freebsd omits the frame pointer, no stack info then -> execinfo segfaults
17:12:42dbussinkMalediktus: ah ok, so you enabled them i guess?
17:12:49dbussinkis this crash you see now reliable?
17:13:15Malediktus-fno-omit-frame-pointer did the trick
17:13:36Malediktusnope, just so it once, out of maybe 15-20 mspec runs :(
17:13:46Malediktuss/so/saw/
17:20:48dbussinkMalediktus: ah, those are pretty tricky then yeah
17:21:19Malediktusyes. but any idea why it didn't demangle? I thought it should...
17:22:02dbussinkMalediktus: this is about as good as it gets with execinfo afaik
17:22:13dbussinkfor a backtrace when it crashes that it
17:22:27MalediktusI thought the vm would use the cxx demangle lib
17:23:48dbussinkdoes seem it tries to do something like that
17:25:13dbussinkMalediktus: but i don't know too much about c++ demangling, if you do, we welcome improvements :)
17:25:37Malediktussure :)
17:25:45Malediktusjust got that crash again
17:31:08dbussinkMalediktus: you can run mspec inside gdb, allows for perhaps catching more information about it
17:37:12Malediktusgood idea
20:22:05TassMalediktus, just set the ulimit for coredumps > 0