summaryrefslogtreecommitdiffstats
path: root/tests
diff options
context:
space:
mode:
authorBernhard Rosenkraenzer <bero@arklinux.ch>2011-03-21 11:19:36 (GMT)
committerOswald Buddenhagen <oswald.buddenhagen@nokia.com>2011-03-21 11:21:12 (GMT)
commite6b61b98f46918d2770a0147a42d10f0a0e18ca8 (patch)
tree099fce2d8b8ade4aa6ec646bb4dba9eb592e77e0 /tests
parent312a46e052dd596d620437f1f5dde0ed5b173c28 (diff)
downloadQt-e6b61b98f46918d2770a0147a42d10f0a0e18ca8.zip
Qt-e6b61b98f46918d2770a0147a42d10f0a0e18ca8.tar.gz
Qt-e6b61b98f46918d2770a0147a42d10f0a0e18ca8.tar.bz2
Fix endianness detection with gcc 4.6 -flto -fwhole-program
Trying to build Qt with gcc 4.6 and the -flto optimization enabled fails at ./configure state because the endianness can't be detected. With -flto in QMAKE_CFLAGS and -fwhole-program in QMAKE_LFLAGS_APP, gcc 4.6 manages to compute msb_bigendian[1] == lsb_littleendian[1] at build time and reduces main() to "return 1;", throwing away the bits being looked for. Treating the short[] arrays as "code" and trying to actually run them prevents even the -fwhole-program optimizer from kicking them out, and since the endian test isn't actually run, doesn't break anything. Merge-request: 1130 Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
Diffstat (limited to 'tests')
0 files changed, 0 insertions, 0 deletions