------- Additional Comments From joseph at codesourcery dot com 2005-05-31 17:05 ------- Subject: Re: [4.1 Regression] 23_containers/bitset/operations/2.cc execution test fails
On Tue, 31 May 2005, pcarlini at suse dot de wrote: > Notice that *nothing* changed in the relevant libstdc++-v3 code. Generally, > only > small, confined changes lately. Either this is a miscompilation or a *very* > long > standing latent problem. Can you attach the *.log excerpt? Even better, take > care > of the obvious checks, like compiling at lower optimization level? (I'm also > going to do that, anyway: we want to ascertain the miscompilation and > recategorize > as soon as possible, in case) Executing on host: g++ -g -O2 -D_GLIBCXX_ASSERT -fmessage-length=0 -DLOCALEDIR="/scratch/gcc/nightly-2005-05-31-mainline/i686-pc-linux-gnu/build_gcc/install/bin/../share/locale" -I/home/gcc/nightlies/src-mainline-2005-05-31/gcc-mainline/libstdc++-v3/testsuite testsuite_abi.o testsuite_allocator.o testsuite_character.o testsuite_hooks.o /home/gcc/nightlies/src-mainline-2005-05-31/gcc-mainline/libstdc++-v3/testsuite/23_containers/bitset/operations/2.cc -include bits/stdc++.h -lm -o ./2.exe (timeout = 300) PASS: 23_containers/bitset/operations/2.cc (test for excess errors) Setting LD_LIBRARY_PATH to :/scratch/gcc/nightly-2005-05-31-mainline/i686-pc-linux-gnu/build_gcc/install/lib:/usr/local/tools/gcc-3.4.0/lib 2.exe: /home/gcc/nightlies/src-mainline-2005-05-31/gcc-mainline/libstdc++-v3/testsuite/23_containers/bitset/operations/2.cc:33: bool test02(): Assertion `b.count() == 0' failed. FAIL: 23_containers/bitset/operations/2.cc execution test (this is installed compiler testing, hence the short command line). Testing manually it seems to fail with -O2 or -O but not -O0. In general when reporting bugs shown up by my regression tester I don't know what part of the compiler they are in; I check enough to make sure they don't appear to be system problems and aren't already reported with reference to the named testcase or discussed on the lists as fixed, but not to determine the exact cause. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21842