Path: utzoo!telly!attcan!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!tut.cis.ohio-state.edu!micro.ti.com!ing-simmons
From: ing-simmons@micro.ti.com
Newsgroups: gnu.gcc.bug
Subject: (none)
Message-ID: <8912150110.AA04160@micro.ti.com>
Date: 15 Dec 89 01:10:50 GMT
Sender: daemon@tut.cis.ohio-state.edu
Distribution: gnu
Organization: GNUs Not Usenet
Lines: 40


  MSG M#=  2001746 FR=NIK  TO=ST2S   SENT=12/14/89  11:04 AM  

There seems to be snag in re-targeting to a processor
which does not define FRAME_GROWS_DOWNWARD but attempts
to set ARG_POINTER_REGNUM and FRAME_POINTER_REGNUM equal.
  In this case FIRST_PARM_OFFSET should be the frame size
of the function - but get_frame_size() returns zero at
this point.
 
   high address
 
    | arg n   |
    .         .
    | arg 1   |
    +---------+ <---- sp at start of prolog +
    | local n |                             |
    .         .                              > Frame size
    | local 1 |                             |
    +---------+ <---- fp                    +
    | reg   n |
    .         .
    | reg   n |
    +---------+ <---- sp at end of prolog
 
   low address
 
This shows up in the (incomplete) 88000 version and
my own experimental version.
 
Am I missing something? or do I need an explicit argument
pointer?
 
(Reason for not defining FRAME_GROWS_DOWNWARD is that target
does not support negative immediate offsets.)
 
Is  bug-gcc@prep.ai.mit.edu  the right address for re-targeting
questions ?
 
Nick Ing-Simmons   ing-simmons@micro.ti.com