Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!tut.cis.ohio-state.edu!HARVARD.HARVARD.EDU!ima!johnl
From: ima!johnl@HARVARD.HARVARD.EDU (John R. Levine)
Newsgroups: gnu.gcc.bug
Subject: code gen bug in gcc 1.34 on 386
Message-ID: <8906060253.AA29810@ima.IMA.ISC.COM>
Date: 6 Jun 89 02:53:07 GMT
Sender: daemon@tut.cis.ohio-state.edu
Distribution: gnu
Organization: GNUs Not Usenet
Lines: 19

The following short routine makes gcc 1.34 on a 386 call abort().  It
compiles OK on a 68000.  I note that there is an entry in the template
table for multiplying two chars to get a short, and that the 386 seems
to be the only machine that actually has such a template.  If I comment
out the template, gcc generates correct code that isn't much worse than
the optimal; maybe it'd be easier to take out the template.

John Levine, johnl@ima.isc.com or Levine@yale.edu

PS:  Did anybody ever do anything with the subset Fortran parser that I
sent in?  I have a better version that I'll send along sooner or later.

f()
  {   
  short foo;
  unsigned char  bar;

  foo = (bar * 3);
  }