new FAILs on HEAD

[Daniel Berlin]

On Sun, 2005-03-13 at 09:45 +0100, Michael Ritzert wrote:
> If you directly receive this mail, your name is tagged to one of the ChangeLog
> entries. A copy is also mailed to gcc-regressions.
> 
> The build system:
> Fedora Core release 3 (Heidelberg)
> i686
> i686-pc-linux-gnu
> configure flags: --enable-threads=posix --prefix=/opt/gcc/install/HEAD --enable-__cxa_atexit --with-system-zlib --enable-languages=c,ada,c++,f95,java,objc
> 
> These new FAILs appeared since the last run:
> 
> FAIL:	cdd2001 - this test appears bogus!
No clue what this even means.

> FAIL: gcc.dg/tree-ssa/20031015-1.c scan-tree-dump-times V_MAY_DEF 2

Grrr, this is me, but never showed up in prior bootstrap + regtest runs
on i686, i looked back through my testresults  to  double check.

We have 3 V_MAY_DEF's now and we should only have two.
The current situation isn't actually incorrect (we may-def both x and
it's subvars, instead of just it's subvars), it's just non-optimal :)

I'm bootstrapping a fix now.