[OpenAFS] rc5 "spvrPort" build problem on i386 FC3....
Jeffrey Hutzelman
jhutz@cmu.edu
Wed, 28 Sep 2005 12:40:23 -0400
On Wednesday, September 28, 2005 10:17:40 AM -0400 Derrick J Brashear
<shadow@dementia.org> wrote:
> On Wed, 28 Sep 2005, Derek Atkins wrote:
>
>> Jeffrey Altman <jaltman@secure-endpoints.com> writes:
>>
>>> Looking at my copy of rxstat_clear_peer.c from rc5 I see no reference to
>>> spvrPort at line 60.
>>>
>>> *srvrPort = strtol(*(argp++), NULL, 0);
>>> if (*srvrPort <= 0 || *srvrPort >= 65536)
>>> Usage();
>>>
>>> Please check to see that this is not local corruption.
>>
>> It is not. The copy of the tarball I pulled down yesterday to start
>> building RPMs has this same problem. So if it is "local corruption"
>> it happened in the creation of the tarball.
>
> ding ding ding ding
>
> we have a winner
>
> I will let Jeff Hutzelman comment further. rc6 (rc5 without this error)
> will follow.
I thought I did comment further, but maybe that was only in another forum.
The tarballs for rc5 were prepared based on a bad checkout, apparently
resulting from a single-bit error in a buffer cache page holding a portion
of the RCS vault file corresponding to the corrupted source file. Since
the checkout contained the corrupted source, so did all the tarballs and
diffs generated from it.
Unfortunately, the machine in question seems to be haunted. No amount of
replacing components seems to be able to fix it. We're in the process of
moving services elsewhere, but it's slow going, because there aren't a lot
of cycles to spend on it.
-- Jeff