------- Comment #3 from jason at gcc dot gnu dot org 2008-08-01 20:01 ------- Assuming that (set SP otherreg) resets args_size to 0 seems strictly better than the current situation, and if it turns out to be wrong we ought to see another crash like this one, so it seems fair to try. Otherwise we'd need to hook into CSE, I guess.
-- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36998