Improve vxWorks support on x86

[Zack Weinberg]

On Fri, 22 Jan 1999 10:50:45 -0800, Mike Stump wrote:
>> Date: Fri, 22 Jan 1999 00:36:57 -0700
>> From: Jeffrey A Law <law@hurl.cygnus.com>
>
>> Mostly OK.  However, you have several ANSI namespace violations that need
>> to be addressed.
>
>> While all the x86 files are not up to date re: these conventions, we
>> are not going to keep adding new violations.
>
>Ok, then we should be willing to fix the current violations?  If so,
>the patch for that is below.  The reason to fix this, is quite a few
>people copy the structure and conventions of existing code, and the
>only way to improve things, is to be willing to remove `bad' examples
>from the code.  We can experiment with the fallout on x86 for a while,
>and after a while, if we like the results, we (I) can come back and do
>the entire rest of the compiler.  I did the edits with an emacs macro,
>so technically they should be high quality.  I avoided the few that
>were already protected or in comments by hand.

Aigh!  This won't work.  CPP_PREDEFINES is interpreted by cpp, not the
driver.  It also won't make any noticeable difference even if it did
work, since the driver always (as far as I can tell) passes cpp the
-undef switch and sets all the predefines itself.

I'd suggest throwing CPP_PREDEFINES away entirely.  If you invoke
gcc-lib/.../cpp directly, you don't get any predefs.  No one is
supposed to do that anyway.  (We could install that shell script
someone posted that maps cpp to gcc -E in $bindir.)

zw