i91@nikhefh.hep.nl (Fons Rademakers) (11/27/88)
I found a peculiar ftn compiler bug (see transcript below).
I challenge everybody to find a smaller program that makes the
ftn compiler barf. :-)
##[2] cat ll.ftn
PROGRAM TT
A=1.0
T=A*1.E10
PRINT *, T
END
##[3] tb
No traceback available
##[4] ftn ll
no errors, no warnings in TT, Fortran version 9.75 1988/11/26 21:55:21 SWIS (Sat)
##[5] tb
No traceback available
##[6] ftn ll -cpu 3000
no errors, no warnings in TT, Fortran version 9.75 1988/11/26 21:55:28 SWIS (Sat)
##[7] tb
floating point operand error (OS/fault handler)
In routine "UID 3FD3DADB.20004950" line 14384
Called from "FLOATOP" line 4087
Called from "PEEPHOLER_OPS" line 4586
Called from "PEEPHOLER" line 5035
Called from "CODEGEN" line 3274
Called from "DRIVER" line 868
Called from "FTN" line 69
Called from "PM_$CALL"
##[8] ll.bin
1.0000000E+10 <------ result seems OK.
Reducing the program to
PROGRAM TT
T = A * 1.E10
END
causes the same error.
The error disappears when
T = A * 1.E9 or less.
The error apeared on a DN580-T as well as on a DN3000.
Anybody wants to comment about this?
-- Fons Rademakers
--
Org: NIKHEF-H, National Institute for Nuclear and High-Energy Physics.
Mail: Kruislaan 409, P.O. Box 41882, 1009 DB Amsterdam, the Netherlands
Phone: (20)5925018 or 5925003 Telex: 10262 (hef nl)
UUCP: i91@nikhefh.hep.nl BITNET: nikhefh!i91@mcvax.bitnetkrowitz@RICHTER.MIT.EDU (David Krowitz) (11/28/88)
The trace back that lists the error message is a traceback of what was left on the stack after the successful completion of the Fortran compiler, not a traceback of the compiled program. I would suspect that the Fortran compiler is not cleaning up the stack properly when it exits, even though it reported no errors during the compilation. -- David Krowitz krowitz@richter.mit.edu (18.83.0.109) krowitz%richter@eddie.mit.edu krowitz%richter@athena.mit.edu krowitz%richter.mit.edu@mitvma.bitnet (in order of decreasing preference)