Weird xcpu2 behavior....

[Eric Van Hensbergen]

Anyone else see anything like this? Looks like random corruption
somewhere in the control stream between xrx and xcpufs....

[root@arlx050 ~]# sphinx/xcpu2/utils/xrx 9.3.61.111 /bin/date
Thu Mar 12 10:48:33 CDT 2009
[root@arlx050 ~]# sphinx/xcpu2/utils/xrx 9.3.61.111 /bin/date
namespace: invalid operation 9Of~
xnamespace failed: : No child processes
[root@arlx050 ~]# sphinx/xcpu2/utils/xrx 9.3.61.111 /bin/date
namespace: invalid operation #9PhA
xnamespace failed: : No child processes
[root@arlx050 ~]# sphinx/xcpu2/utils/xrx 9.3.61.111 /bin/date
namespace: invalid operation Î0A
xnamespace failed: : No child processes
[root@arlx050 ~]# sphinx/xcpu2/utils/xrx 9.3.61.111 /bin/date
namespace: invalid operation w§ºQ
xnamespace failed: : No child processes
[root@arlx050 ~]# sphinx/xcpu2/utils/xrx 9.3.61.111 /bin/date
Thu Mar 12 10:48:44 CDT 2009

[Latchesar Ionkov]

I think the reason is a bug in the ns file parsing code. I thought I
fixed it, but may be it reappeared again. I'll look into it in the
afternoon.

Thanks,
Lucho

[Abhishek Kulkarni]

I can't reproduce this anymore. But does the following patch from
http://www.mail-archive.com/xc...@googlegroups.com/msg00068.html
fix the issue for you?

Index: xcpufs/xcpufs.c
===================================================================
--- xcpufs/xcpufs.c     (revision 691)
+++ xcpufs/xcpufs.c     (working copy)
@@ -1332,7 +1332,7 @@
                s = p;
        }

-       bufwrite(&buf, buf.size, strlen(s), s);
+       bufwrite(&buf, buf.size, (slen + str - s), s);
        b = 0;
        bufwrite(&buf, buf.size, 1, &b);

[Eric Van Hensbergen]

That does seem much more stable...

       -eric

[Latchesar Ionkov]

Do you still see problems with that change?

Thanks,
Lucho

[Eric Van Hensbergen]

I haven't seen the problem since I patched and that includes a demo,
so it's passed a murphy's law barrier as well.

Sent from my iPhone

[Abhishek Kulkarni]

That patch got buried in the queue somewhere. Should I take this as an ack?

[Latchesar Ionkov]

Acked-by: Latchesar Ionkov <lu...@ionkov.net>

[Eric Van Hensbergen]

I just tripped over this in a fresh checkout of the xcpu2 code from
sourceforge, whose responsible for comiting the patch now that its
ack'd.

-eric

[Abhishek Kulkarni]

On Tue, Mar 31, 2009 at 1:42 PM, Eric Van Hensbergen <eri...@gmail.com> wrote:

I just tripped over this in a fresh checkout of the xcpu2 code from
sourceforge, whose responsible for comiting the patch now that its
ack'd.

My bad, I didn't see the ack at all.
Thanks for the gentle reminder, r758.