|Summary:||kill -QUIT produces a very broken stacktrace|
|Product:||[Mono] Runtime||Reporter:||Alan McGovern <alan>|
|Component:||Debugger||Assignee:||Zoltan Varga <vargaz>|
|Severity:||major||CC:||david.karlas, lluis, mono-bugs+mono, mono-bugs+runtime, shrutis, vargaz|
|Tags:||Is this bug a regression?:||---|
|Last known good build:|
|Bug Depends on:|
Description Alan McGovern 2016-02-07 12:01:22 UTC
Neither mono's `kill -QUIT $PID` method of writing stacktraces nor gdb's `t a a bt` showed that managed code was running in the main thread. Only LLDB did. Can we at least fix `kill -QUIT` because we rely on that *a lot* to figure out what our threads are doing. Here is an example threaddump from mono, gdb and lldb where only lldb shows managed code in the main thread. https://gist.github.com/alanmcgovern/9eee935fc26efbe3655e
Comment 1 Zoltan Varga 2016-02-07 20:47:58 UTC
This is probably caused by the monitor changes, we directly call the monitor_enter () etc. c functions without a wrapper so we can't produce an async stacktrace if they block.
Comment 2 Zoltan Varga 2016-02-11 02:30:37 UTC
Fixed in mono master c5512c0d3e2b102ed16158ce8c599c8203ccd510 and the mono-4.3.2-branch 9f44a62acf718d91320420c1c77e855a63f3c3f9.
Comment 3 Shruti 2016-06-02 11:25:40 UTC
@Alan, @Zoltan : It would be great if you let us know that what and how to reproduce this issue. This issue must be reproduced at our end before verifying it. Looking for your help! Thanks!!
Comment 4 Zoltan Varga 2016-06-02 12:36:10 UTC
Its not a product problem, it doesn't need to be verified.
Comment 5 Shruti 2016-06-02 13:19:48 UTC
Thanks @Zoltan for the information. I am closing this issue as per comment(4).