I am not a member of this list, so please do not expect me to see
replies to this message on this list.  I wanted to report a problem that
we encountered with NFS on RedHat 6.0 (kernel 2.2.5).  We encountered a
problem with somewhat random error when compiling code on a disk mounted
via NFS to a Sun NFS server running Solaris 2.6.  I say the problem was
random because it didn't happen very often, but it was somewhat
repeatable, in that compilation would usually fail on the same files.

The problems that occurred were sometimes compiler crashes and sometimes
errors in linking.  The system that we were building would build fine if
we built it on a local disk or if we built it on an NFS disk resident on
an SGI server, but would consistently fail on the Sun served disk.  It
would also compile fine on a RedHat 5.2 system (on the same
filesystem).  I examined one file that would consistently fail, which
seemed to be generating a corrupted object file, and the files differed
by 4 bytes.  Two nulls (^@) were inserted in one part of the file and
two nulls were missing from another part of the file.  I tried copying
large amounts of data between the various disks/systems and never saw
any differences in the file, so I suspect the problem has something to
do with the way gcc is writing the data to the object files.

We were apparently able to solve the problem by setting the wsize and
rsize parameters to 1024 (at least I haven't seen any more problems
since the change), but I wanted to make sure people are aware there
seems to be a bug somewhere.  This could also be a but in Solaris 2.6,
but it only shows up with RedHat 6.0 clients.

Let me know if anyone would like more information.

Brian Webb
[EMAIL PROTECTED]

begin:vcard 
n:Webb;Brian
tel;work:(520) 570-7649
x-mozilla-html:FALSE
version:2.1
email;internet:[EMAIL PROTECTED]
adr;quoted-printable:;;SAIC=0D=0A101 N. Wilmot  Suite 400=0D=0A;Tucson;AZ;85711;USA
x-mozilla-cpt:;32008
fn:Brian Webb
end:vcard

Reply via email to