Possible Bug

[Jim Bean]

hi,
I started getting reports of a driver failing after folks upgraded their
systems.  They were all
now using gcc version  2.91.66  (RH 6.0 and SUSE 6.1).  I verified this
locally.    I haven't
sent you much documentation as this may be a known/fixed problem.  If
not I will gladly send
a more complete package.
The problem occurs when reading a hardware register from an addin A/D
converter board using inb_p.  Some bits in the register are tied
directly to hardware switches.  These bits can not change during
execution.  The register is read in two places and the bits appear to
change.  The register is also written to.  It seems as if the register
value is being cashe'd.
I have tried optimization levels O1, O2 and none, and get the same
result.
I will try using the latest version of GCC if necessary, but the machine
that shows the failure is not "netable" so this would not be easy to
do.  Enclosed are the driver C code and a dmesg dump showing the result
of the two reads.
regards
Jim Bean