------- Comment #12 from janus at gcc dot gnu dot org 2009-09-10 18:13 ------- (In reply to comment #10) > But I also think that maybe fixing the code so that the double resolve is no > harm in this case is the better way to go; something like adding a flag that > code is from PPCs and shouldn't be resolved and returning early, if there's no > other solution... What do you think?
I think it would be a better practice to resolve everything exactly once (to avoid unnecessary overhead), and it seems that most of resolve.c does stick to this rule (although I'm not completely sure about that). At least I had no problems before with my assumption that PPCs would be resolved just once. Also I'm not sure if multiple resolution could hurt in other places where it is not expected. Moreover, in this case it is just wrong: Calling resolve_ordinary_assign should just resolve the assignment statement, and not the stuff that comes after it (here: the PPC call). -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41242
