[Mono-dev] mono crash when multiple threads tries to lock files on Linux

[Rodrigo Kumpera]

File a single bug and attach all file as a single zip one.

Thanks for the time of figuring out a test that reproduces it.

On Sun, Sep 18, 2011 at 8:36 AM, Esben Laursen <el at rantek.dk> wrote:

> Hi All,
>
> I have a problem that then I try and open close the same file multiple
> threads it eventually crashes with a SIGSEGV.
>
> The problem that I have in my app actually happens on the Dispose method,
> but I have managed to write a repro  that fails when on the open (I do not
> know if they are related, but it still crashes so it must also be a bug).
>
> With mono-2.6.7 it does not happen after a while (depending on how many
> threads that are started), and the stack trace is not very informative, but
> with mono-2.10.5 it happens almost  immediately and the stacktrace gives a
> lot more output.
>
> My Question, do you guys think it would be best to create 2 bug reports,
> one for 2.10.x and one for 2.6.x branch or would it be better to just create
> one?
>
> If someone is interested I have uploaded the files in question to my
> dropbox if someone are interested..
>
> I have compiled the source with "mcs -debug+ FileLockRepro_monoDevelop.cs"
> or debugged it directly from MonoDevelop.
>
> http://dl.dropbox.com/u/2752232/FileLogRepro/FileLockRepro_monoDevelop.cs
>
> http://dl.dropbox.com/u/2752232/FileLogRepro/mono-2.6.7-orginal-stacktrace.txt
>
> http://dl.dropbox.com/u/2752232/FileLogRepro/mono-2.6.7-repro-output_and_stacktrace-50threads.txt
> http://dl.dropbox.com/u/2752232/FileLogRepro/mono-2.6.7-versions.txt
>
> http://dl.dropbox.com/u/2752232/FileLogRepro/mono-2.10.5-repro-output_and_stacktrace.txt
>
> http://dl.dropbox.com/u/2752232/FileLogRepro/mono-sgen-2.10.5-repro-output_and_stacktrace.txt
> http://dl.dropbox.com/u/2752232/FileLogRepro/mono-2.10.5-versions.txt
>
>
> Do you think you guys need more information to debug it? As I don't really
> have any idea where to start :-)
>
> Ohh yes, do you think the orginal stacktrace could be compared to the repro
> one?
>
> Many thanks for your help..
>
> Cheers
>
> Esben
> _______________________________________________
> Mono-devel-list mailing list
> Mono-devel-list at lists.ximian.com
> http://lists.ximian.com/mailman/listinfo/mono-devel-list
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.ximian.com/pipermail/mono-devel-list/attachments/20110919/3b33c2a3/attachment-0001.html