Brian Drummond wrote: > As it is, if I read it correctly, function type_for_size (ortho_lang.c > line 690ff) can potentially return an uninitialised tree in cases where > previously it would return NULL_TREE.
I don't see that. The function returns the result of either make_unsigned_type or make_signed_type. Assuming those functions are well-behaved, type_for_size should always return something. I will admit that I don't know exactly what I am doing. The previous code was clearly wrong as evidenced by the internal error on MIPS. I suspect it had something to do with a lack of distinction between floating point and integers. My approach was to look carefully at other language front-ends (Ada, C) and stay closer to their implementation of type_for_size and type_for_mode. Joris. _______________________________________________ Ghdl-discuss mailing list [email protected] https://mail.gna.org/listinfo/ghdl-discuss
