Re: Bug in gcc

[Mark Rivers <RIVERS@cars3.uchicago.edu>]

    Folks,

The problem I was having with medm crashing on a DEC Alpha machine has been
solved.

The fundamental problem was the bug in Channel Access in R3.12.2 which Jeff
Hill just described.  The low-order 32 bits of doubles were not being
byte-swapped correctly.

The contributory problem was the bug in gcc which I described last week.  gcc
2.2.3.1 is generating -0.0 values rather than 0.0 in certain circumstances.
-0.0 is a legal floating point value (though not what was intended).  However,
when -0.0 is passed through Channel Access with the byte-swap bug, then it
results in an invalid (denormalized) floating point number, which crashes medm.

When trying to track down the problem I built all of R3.12.2 with a more
recent version of gcc, 2.5.8.  This version does not have the -0.0 bug. EPICS
base compiled fine, but crashed with an access fault in a device
initialization routine before iocInit.  Is anyone running EPICS compiled on a
more recent version of gcc? 2.2.3.1 is more than 4 years old, and it is likely
that other bugs besides this one have been fixed since then.

____________________________________________________________
Mark Rivers                             (773) 702-2279 (office)
CARS                                    (773) 702-9951 (secretary)
Univ. of Chicago                        (773) 702-5454 (FAX)
5640 S. Ellis Ave.                      (708) 922-0499 (home)
Chicago, IL 60637                       rivers@cars.uchicago.edu (e-mail)

or:
Argonne National Laboratory             (630) 252-0422 (office)
Building 434A                           (630) 252-0405 (lab)
9700 South Cass Avenue                  (630) 252-1713 (beamline)
Argonne, IL 60439                       (630) 252-0443 (FAX)