Just upgraded to 1.4 rc1 and ruby won't build. Could this be a gcc3.2 issue? chmod 0644 /mnt/zeta/var/tmp/portage/ruby-1.6.7/image//usr/lib/ruby/1.6/irb/ruby-token.rb lib/irb/completion.rb -> /mnt/zeta/var/tmp/portage/ruby-1.6.7/image//usr/lib/ruby/1.6/irb/completion.rb /mnt/zeta/var/tmp/portage/ruby-1.6.7/work/ruby-1.6.7/lib/ftools.rb:33: [BUG] Segmentation fault ruby 1.6.7 (2002-03-01) [i686-linux-gnu] make: *** [install] Aborted !!! ERROR: The ebuild did not complete successfully. !!! Function src_install, Line 1, Exitcode 2 !!! (no error message)
I cannot reproduce this. Tested it with 1.4_rc1, gcc-3.2, CFLAGS="-march=i686 -O3 -pipe", and CFLAGS="-mcpu=pentium -O3", for dev-lang/ruby-1.6.7, even using nothing but the i386-stage3 tarball. Unless it stops at the same place every time you compile it, it's most likely a hardware problem, such as faulty RAM. It may also be a problem with too aggressive optimisation options; try using just -O3. Without being able to reproduce this, there's little I can do to fix it. If you figure out what causes this, please reopen the bug and I'll have another go.
The exact problem occurs at the exact same place for me as well. Just to ensure it is not a memory problem, I have run it 10 times in a row, each failing at the exact same spot. I have a very unagressive set of CFLAGS.... CFLAGS="-march=athlon -O3 -pipe -fomit-frame-pointer" HOWEVER, if you remove -fomit-frame-pointer it compiles cleanly. I am running 1.4 fully updated on a regular basis. I suspect given the flag change causes it to work, this is a GCC 3.2 issue.
As far as I see it you can do one of 3 things: 1. Ignore this. 2. Put an 'ewarn' in the ebuild stating that an over-aggressive CFLAGS will cause compilation problems. 3. Actually checking CFLAGS and removing "bad" options. #2 seems the easiest. If the compilation fails, the user will see the warning messages and know to adjust their CFLAGS. Perhaps putting that you think the problem is -fomit-frame-pointer would be helpful.
Perhaps the ruby team should have a list of sanctioned optimisation flags? I can't get to this anytime soon, I'm afraid.
*** Bug 12545 has been marked as a duplicate of this bug. ***
Ebuilds updated to use flag-o-matic/filter-flags. -fomit-frame-pointer was already filtered out when I update this so presumably it's already fixed? Will keep an eye out for other CFLAGS that cause problems.
