[tuples] Fix gimplification of MODIFY_EXPR

[Diego Novillo]

These are a few miscellaneous fixes found while trying to compile libgcc.

- We were not consistently checking what is allowed on the RHS of an
assignment.  I reformulated is_gimple_formal_tmp_rhs in terms of a new
predicate that returns whether a tree expression is a gimple binary,
unary or single object RHS.

- The code for attaching statements to sequences was allowing sequences
that were "cut in the middle", added new asserts to make sure that
sequences are cut off at the ends.

Tested with gimple.exp.  Committed.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 20070727-various-fixes.diff
Type: text/x-patch
Size: 12774 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20070728/ef75affe/attachment.bin>