This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [libobjc] fix to nil_method
> Hmmm, interesting. I'm guessing that the bug is probably in how ObjC
> semantics produce the call to nil_method - going by past experience,
> it's probably being sloppy with decl nodes somewhere and not preserving
Quite probably.
> variadicity properly. In theory, I could try to debug by comparing
> cross compiler output, but I'm not sure what the desired code should
> look like (your patch tweaking library internals rather than the
> test program).
The resulting call should have extra xor %rax, %rax instrution in it
to set number of FP instructions to 0.
I can debug it after the vacation (till 29 september), but another
question if effeciency. My patch just removes dummy initialization
of reg_save_area code, so resulting code is much more effecient
(I am not sure if that matters).
Is there some target in existence where calling nonvariadic function
with variadic prototype breaks?
Honza
>
> Stan