From mboxrd@z Thu Jan  1 00:00:00 1970
From: Zack Weinberg <zack@wolery.cumb.org>
To: John David Anglin <dave@hiauly1.hia.nrc.ca>
Cc: gcc-patches@gcc.gnu.org, gcc-bugs@gcc.gnu.org
Subject: Re: VAX Ultrix bootstrap failure with gcc-2.96
Date: Thu, 25 May 2000 19:35:00 -0000
Message-id: <20000525193543.W3314@wolery.cumb.org>
References: <20000525183119.T3314@wolery.cumb.org> <200005260220.WAA11033@hiauly1.hia.nrc.ca>
X-SW-Source: 2000-05/msg01435.html

On Thu, May 25, 2000 at 10:20:54PM -0400, John David Anglin wrote:
> > It did catch some real bugs, so I'm not complaining.  What are the
> > $ in identifier messages complaining about?  It's probably another
> > warning that should be turned off in system headers, or something like
> > that.
> 
> The $ in identifier messages are complaining about lines like
> 
> #if     defined(__GFLOAT) || CC$gfloat
> 
> which occur in limits.h (syslimits.h) and math.h in the gcc include
> directory.  The identifier CC$gfloat is something the VAX ANSI C
> compiler defines.  Probably, this should be nuked when the includes
> are fixed.  It obviously isn't needed in the gcc include directory.

I agree.  Could you come up with an entry for inclhack.def?  It will
be easier for you since you have access to the header files.  I'll
guess at an appropriate pattern:

fix = {
  hackname = vax_math_ifdefs;
  files = limits.h;
  files = math.h;
  select = "^#if.*||[ \t]+CC$[a-z]+";
  sed = "/^#if/s/||[ \t][ \t]*CC$[a-z][a-z]*//g";
};

This will undoubtedly need refinement to do the right thing in all cases.

> Another small related issue is that gcc should define __GFLOAT
> in addition to GFLOAT when the -mg option is used.  I have edited
> my specs file to do this.

Patch for config/vax/*.h, please?

zw