[Mono-devel-list] Mono segfaults with updated kernel?? 2.6.10

[jrodman at mono-list.spamportal.net]

On Fri, Jan 28, 2005 at 01:27:53PM +0100, Paolo Molaro wrote:
> On 01/19/05 Joshua Rodman wrote:
> > For those who might be interested, attached is the output of strace32.
> > It seems like the process is dying during initial load/link.  This is
> > running '/usr/bin/mono /usr/lib/mono/1.0/mcs.exe' with no arguments.
> > Am I correct?  Is my distribution broken?
> [...]
> > open("/usr/local/NUnit/bin/mscorlib.dll", O_RDONLY|O_LARGEFILE) = -1 ENOENT (No such file or directory)
> > open("/usr/lib/mscorlib.dll", O_RDONLY|O_LARGEFILE) = 4
> 
> This is most likely an old mscorlib.dll: in a correct default install
> it should load mscorlib.dll from /usr/lib/mono/1.0/.
> Try removing it.

surfboard:~ # rpm -qf /usr/lib/mscorlib.dll
mono-core-1.0.5-1.novell.1.4

Are the rpms supplied by Ximian/Mono incorrectly built?

> The other option is that you might have and amd64 processor and the
> new kernel enables the exec protection by default, so you'll need
> to use mono svn for that fix (or wait for 1.0.6).
> Or, since the stacktrace you posted starts from mono_domain_get(), it 
> may be that tls access is broken on your system.

Ahhh.  Apparently this is the problem.  
Forgive my sour grapes in saying that I explicitly stated my problem was
on an AMD Opteron in the first report of this problem.

Thanks for suggesting it.

I will now investigate reading through the SVN changelogs to find the
specific patch.

-josh