On Tue, Jun 24, 2014 at 11:16 PM, Eric Botcazou <ebotca...@adacore.com> wrote: >> I think that was on purpose to avoid arithmetics in enum types. As those >> conversions are useless and thus stripped later is it really important >> to retain enum and boolean kind here? > > The problem is that convert.c is called by front-ends and the patch also > removed the callback into them that made it possible to have some control. > So, yes, it's pretty annoying to see totally bogus conversion nodes being > introduced into your ASTs behind your back...
Ok, but the (previous and proposed) behavior is odd as for unsigned_type_for (boolean_type_node) you'd get back a BOOLEAN_TYPE but for unsigned_type_for (signed_type_for (boolean_type_node)) you get back an INTEGER_TYPE. That is, because we use build_nonstandard_integer_type now the result will always be an INTEGER_TYPE (unless you happen to pass in a type with the correct signedness with your patch). I don't like re-introducing that inconsistency. Maybe instead make convert.c do if (!TYPE_UNSIGNED) unsigned_type_for () instead? I notice that all callers of [un]signed_type_for are in "premature" optimizations convert.c performs (that better should be done in fold-const.c). Thanks, Richard. > -- > Eric Botcazou
