Re: exceptions and m68k coff

[Jason Merrill <jason at cygnus dot com>]

>>>>> Philippe De Muyter <phdm@macqel.be> writes:

> You'll find below the assembler output by egcs-971031 when compiling _muldi3
> with -fexceptions (with and without -dA).  The assembler complains on it with :
> Assembler: libgcc2.c
>  aline 158 : Fatal: illegal subtraction
>  aline 175 : Fatal: illegal subtraction

> Not all the substration are illegal, only the one involving forward
> references.  I understand those substraction are there to compute the
> size of the entries.  Would it not be easy to compute them in gcc itself,

Not easy, but certainly possible.  It used to work that way, actually.  The
idea is for it to work that way again.

> I also surmise other problems : I have compiled two times : with and
> without -dA.  From comments I've read, I expected to find "eh" written
> somewhere. Is that to be considered also as a bug ?

No, you only get "eh" for translation units that have exception regions.

> I also see byte and longs mixed in output data.  Do the functions using
> those entries expect to find the longs aligned or unaligned (we are on a
> m68k here).

unaligned.

> We are on a coff system here.  My OS only allows to load 3 sections :
> .text, .data, and .bss.  Should the frame info not go in the text section
> ?

No, data.

Jason