[tuples] Fix gimplification of COND_EXPRs

[Diego Novillo]

Chris found this last week and the type verifier tripped over it as
well. When gimplifying COND_EXPRs we were switching around the true
and false labels.  Also, when gimplifying a ?: conditional, instead of
generating GIMPLE for the canonicalized conditional jump, we were
generating GIMPLE for the original expression, which was causing some
type mismatches when the conditional expression involved pointers.

This was causing a failure in gcc.c-torture/compile/struct-non-lval-2.c
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 20070830-fix-cond_expr-gimplification.diff
Type: text/x-patch
Size: 1110 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20070830/f256cfab/attachment.bin>