On Fri, Sep 12, 2003 at 12:30:03PM +0200, Michael Wood wrote:
> > This test seems to confirm the theory of a bug/race condition in the
> > nfs code. Nevertheless, I would like to track down the nfs problem.
>
> Maybe linux-kernel would be interested. Especially if you can write a
> small test case that demonstrates the problem.
I will propably do this next week, when the system administrator of the nfs
server is back from vacation (and I can get all the details of the server
setup).
But now I can at least confirm, that the problem not subversion related.
I wrote a small C program which executes the sequence
rename("old/foo","new/foo");
stat("new/foo",tmpstat);
chmod("new/foo",S_IRUSR);
rename("new/foo","old/foo");
in an endless loop; after 50 to 1000 iterations, the same error occurs
when calling chmod(). Btw, the error seems only to occour, if one renames
accros directories, e.g. a rename("old","new") could not provoke the
race condition. But I will gather more information about our setup and
report it to the kernel list next week.
Thanks for your help!
Achim
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Sep 12 14:50:12 2003