f90-intrinsic-bit.f (from the gfortran test suite) fails with at any
optimization level different that -O0. This is due to the flag
-fmerge-constants. The reduced test case is the following:
integer*1 k
k = 0
print *, ISHFT(k,-BIT_SIZE(k))
end
At compilation, it produces:
$ gfortran -fmerge-constants reduced.f
/usr/bin/ld: /var/tmp//cceQRkH8.o 8 byte literal section (__TEXT,__literal8)
size is not a multiple of 8 bytes
The assembler output with and without differ only by one line. With the flag on
(wrong assembler), it is:
LC0:
.ascii "reduced.f\0"
.literal8
while the "correct" assembler (without -fmerge-constants) is:
LC0:
.ascii "reduced.f\0"
.const
I hope this will be enough for you to identify the source of this bug, which is
currently the only test case failure specific to powerpc-darwin...
--
Summary: f90-intrinsic-bit.f produces wrong code with -fmerge-
constants on powerpc-apple-darwin7.7.0
Product: gcc
Version: 4.0.0
Status: UNCONFIRMED
Severity: normal
Priority: P2
Component: fortran
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: coudert at clipper dot ens dot fr
CC: gcc-bugs at gcc dot gnu dot org
GCC build triplet: powerpc-apple-darwin7.7.0
GCC host triplet: powerpc-apple-darwin7.7.0
GCC target triplet: powerpc-apple-darwin7.7.0
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19335
