Re: flexfax: Race conditions in faxq?

[Marisa Giancarla <marisa@andromedia.com>]

>>>>> "Gabriel" == Gabriel Fernandez 4-6518 <fernande@mr.med.ge.com> writes:

    Gabriel> Hi Marisa: This might be a short in the dark, but the
    Gabriel> instruction set in the CPU may have changed and is not
    Gabriel> 100% backward compatible. You might have to recompile the
    Gabriel> whole thing again to make it work.

	Some, exact same type of CPU, just faster speed and
surrounding hardware. Additionally, it WAS recompiled when the problem
first popped up with the exact same results.

    Gabriel>    Also, you mention you changed machines, what about the
    Gabriel> OS? Did it change too? This may have chenged the way I/O
    Gabriel> is handled.

	Nope, exact same revision of the exact same OS.

	As I said, I put some syslog entries in it for debugging which
caused processing of the fax control headers to go a tiny bit slower
and it started working fine. So after I sent the message yesterday I
took a clean dump of the source from my backup and put a single
usleep() call right before the "case" statement which was causing the
fault, and it works fine now. Tried 500us first, didn't help, went to
1ms and it worked, so I went to 1.5ms just to be safe. If that line is
removed the error pops up again, with it back it works. I am sure
there is a cleaner way to fix it than that, but I have too many other
projects to spend time learning the rest of the hyklafax code to see
how...

	Marisa