Created attachment 593620 [details] Test source to produce the error Description of problem: I found that cups-pdf was failing to write pdf files to the users home directories that were housed in a gluster volume. After added some proper error checking to the source, I discovered that it was failing with EOVERFLOW. This affects nearly all EL5 packages. I was able to work around it by defining "-D _FILE_OFFSET_BITS=64" in rebuilding that one file I needed. This can be tested by compiling the attached: gcc -g -O0 -o stat stat.c Copy that to a gluster volume mounted on 32 bit EL5 and run it. It will return: stat: Value too large for defined data type stating stat: -1 00000000a006a2b9 Note the 32bit truncation of the inode. Compiling with: gcc -D _FILE_OFFSET_BITS=64 -g -O0 -o stat stat.c and performing the same test produces stating stat: 0 00000000a006a2b9 Notice it's still a truncated inode (actual inode was 9bd28da4a006a2b9). Since EL5 cannot be rebuilt with _FILE_OFFSET_BITS=64, there should be a solution in the fuse interface.
Joe, with recent patchset to solve bug 850352 fix this issues? http://review.gluster.org/3955 is submitted and merged in master.
This seems to be a duplicate of bug 850352. Please re-open the bug if found otherwise. *** This bug has been marked as a duplicate of bug 850352 ***