https://bugs.kde.org/show_bug.cgi?id=401828
Will Schmidt <will_schm...@vnet.ibm.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Ever confirmed|0 |1
Status|REPORTED |CONFIRMED
--- Comment #1 from Will Schmidt <will_schm...@vnet.ibm.com> ---
Started to investigate this one. Still more investigation to do, but I wanted
to do a snapshot braindump at this point.
It looks like the behavior is affected by the optimization level used to build
the testcase itself. Using a significantly stripped down test that exercises
the fcfids instruction similar to how the valgrind test does, I see
#
$> gcc -v
gcc version 9.0.0 20181130 (experimental) (GCC)
> gcc test_fcfids.c -O0 -o test_fcfids
11:37:31 0 p8 ppc64le
willschm@pike:~/valgrind_builds/valgrind/none/tests/ppc64> valgrind
./test_fcfids
==104410== Memcheck, a memory error detector
<...>
0.000000
0 0
testcase 0000000000000000 => (raw sp) 0 )
1000.000000
1 4.65201e+18
testcase 408f400000000000 => (raw sp) 5e811e80 )
But with -O or -O1 (or higher), the 5e811e80 value disappears.
11:39:04 0 p8 ppc64le
willschm@pike:~/valgrind_builds/valgrind/none/tests/ppc64> gcc
test_fcfids.c -O1 -o test_fcfids
11:39:12 0 p8 ppc64le
willschm@pike:~/valgrind_builds/valgrind/none/tests/ppc64> valgrind
./test_fcfids
==104524== Memcheck, a memory error detector
<...>
0.000000
0 0
testcase 0000000000000000 => (raw sp) 0 )
1000.000000
1 4.65201e+18
testcase 408f400000000000 => (raw sp) 0 )
# but is still present when run outside of valgrind.
11:39:14 0 p8 ppc64le
willschm@pike:~/valgrind_builds/valgrind/none/tests/ppc64> ./test_fcfids
0.000000
0 0
testcase 0000000000000000 => (raw sp) 0 )
1000.000000
1 4.65201e+18
testcase 408f400000000000 => (raw sp) 5e811e80 )
--
You are receiving this mail because:
You are watching all bug changes.
