This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PATCH] PPC: Correct SYSV ABI libcall handling
- From: Franz Sirl <Franz dot Sirl-kernel at lauterbach dot com>
- To: David Edelsohn <dje at watson dot ibm dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Tue, 17 Jun 2003 22:46:31 +0200
- Subject: Re: PATCH] PPC: Correct SYSV ABI libcall handling
- References: <200306172026.QAA24732@makai.watson.ibm.com>
On Tuesday 17 June 2003 22:26, David Edelsohn wrote:
> >>>>> Franz Sirl writes:
>
> Franz> Yes, the doc for the FUNCTION_ARG macro says TYPE is only zero for
> libcalls, Franz> and not even that is always true (only for the first round
> of FUNCTION_ARG Franz> calls). That is also what the old comment above
> function_arg suggested. Franz> So the test for "type" was only there to
> rule out libcalls.
>
> It wouldn't be better to make "type" accurately match the
> documentation?
Well, I've talked with rth on IRC about this, and if I recall correctly that
would be a _very_ invasive change (due to the way FUNCTION_ARG is currently
used) and instead he wanted me to get rid of libcalls altogether and generate
them like any other call. That is, however, totally out of the timeframe I
can spend on GCC. So I went with the simplest solution I could think of.
BTW, if you read the docs closely, the case of TYPE being nonzero for a
libcall is _not_ ruled out:
TYPE, the data type of the argument as a tree node or 0 if that is not known
(which happens for C support library functions).
Comments in other header files suggest the same, eg. alpha.h merely talks
about that TYPE "may" be zero for libcalls.
Franz.