Updates:
Status: Started
Comment #6 on issue 2513 by [email protected]: Patch: Build: add -fno-tree-vrp
to CXXFLAGS for gcc 4.7.0
http://code.google.com/p/lilypond/issues/detail?id=2513
-O1 are "most effective and affordable optimizations", so if -O1 -ftree-vrp
makes bad code already, I agree that -fno-tree-vrp seems like our best bet
at the current point of time.
LGTM. To avoid delays, I would strongly suggest putting this on the same
or an earlier countdown than issue #2514 and pushing issue #2513 first. It
would also be a prime candidate for cherry-picking into 2.16 if it that
gets released as currently planned, and pushing it earlier than issue #2514
will make sure that one can cherry-pick it independently from that issue.
It certainly is more important.
Thanks for the really thorough testing you made here. I also see that you
reported the problem to GCC developers and that they are on track.
Unfortunately it turns out from the discussion that the tree-vrp pass is
not responsible in itself (and as I stated in comment #3, it has been
around for quite a long time) but is working from bad input.
So let us hope that the bug _only_ manifests itself in the work of
tree-vrp, or we will potentially have more fun with the effects of this bug
in other places.
Anybody from those with project administration rights on code.google.com: I
can't make Pavel the owner of this issue. Can you add him to the members
of this project?
