[tree-profiling-branch] Fix mudflap

[Steven Bosscher]

On Friday 21 May 2004 18:05, Frank Eigler wrote:
> Steven Bosscher <stevenb@suse.de> writes:
> > [...]  On the tree-profiling-branch, all mudflap test were failing
> > because mudflap is not a CFG-aware pass.  The attached patch changes
> > this.  It makes mudflap2 work on the CFG.  [...]
>
> Thanks.  I have no objection to putting this on mainline too, if it
> does not worsen the test results.

I don't think it would.  I have fixed one of the 5 remaining failures
and the other four are cases where some c++ headers cannot be found,
so compilation never even reaches the point where mudflap instrumentation
is inserted.  So I think it's completely in working shape now for the
tree-profiling branch.

>  (I don't know if mainline ssa
> already makes it possible for mudflap2 to deal with CFGs instead of
> trees.)

Well, it could be done but it needs a little more re-ordering of the
passes than what's required on the tree-profiling branch.  And since
mainline works as-is, it can wait.

Gr.
Steven