summaryrefslogtreecommitdiff
path: root/src/rpc/genprotocol.pl
diff options
context:
space:
mode:
authorEric Blake <eblake@redhat.com>2011-12-02 11:56:58 -0700
committerEric Blake <eblake@redhat.com>2011-12-02 13:00:57 -0700
commit75da3200876a8b92ab78b15d3772809fef794faf (patch)
tree94edc9192f3013703711abde4218391e64de99f8 /src/rpc/genprotocol.pl
parent4cfdbfc46f0a7b777035a5b2af5ecbe7373a7e12 (diff)
downloadlibvirt-75da3200876a8b92ab78b15d3772809fef794faf.tar.gz
build: fix build at -O2 on rawhide
I had previously tested commit 059d746 with -O intentionally omitted from my CFLAGS; but that means that I missed out on this warning from gcc 4.6.2 when optimizations are enabled: util/buf.c: In function 'virBufferGetIndent': util/buf.c:86:1: error: function might be candidate for attribute 'pure' [-Werror=suggest-attribute=pure] While it is probably a good idea to add the attributes and silence this warning, it's also invasive; 'make -k' found more than 75 such complaints. And it doesn't help that gcc 4.6.2 is still buggy (coreutils reported a case where gcc 4.6.2 incorrectly suggested marking a function pure that incremented a global variable; fixed in gcc 4.7). So the best fix for now is to disable the warning. It also doesn't help that I stumbled across another problem - gcc documents that -Wsuggest-attribute=pure only warns if you use -O, or if you use -fipa-pure-const. But in practice, when I omitted -O but added -fipa-pure-const, the warnings are fickle - I got warnings for simple compilation that disappeared when I also added -fPIC. And the way libtool compiles things is with -fPIC first, then without -fPIC but with errors sent to /dev/null - which meant that without disabling -Wsuggest-attribute=pure, I got a compile error with no message. :( See http://debbugs.gnu.org/cgi/bugreport.cgi?bug=10197 * m4/virt-compile-warnings.m4 (LIBVIRT_COMPILE_WARNINGS): Silence -Wsuggest-attribute warnings for now.
Diffstat (limited to 'src/rpc/genprotocol.pl')
0 files changed, 0 insertions, 0 deletions