[PATCH] Clean-up math builtin argument stabilization

[Richard Henderson]

On Wed, Jun 18, 2003 at 05:34:42PM -0600, Roger Sayle wrote:
> But in this case, we throw away the SAVE_EXPR as well, as its now only
> local to this function.  Because we no longer modify the tree, we fall
> back to the expand_builtin_function which then attempts a library function
> call using the tree that doesn't contain the SAVE_EXPR.

*Still* not safe.  Please do look at calls.c, as I mentioned.


r~