Re: %[spec] and x86 cpp option problems

[Jeffrey A Law <law at cygnus dot com>]

  In message <199807092314.QAA01070@ada.cygnus.com.cygnus.com>you write:
  > The x86 ports have 2 minor bugs in the the cpp option support.
  > 1) When -ansi is used, they define macros in user namespace, e.g. -Di386.
  > 2) Some macros defined by gcc-2.7.2.3 aren't defined anymore, e.g. -D__i386.
Agreed.

  > rarer and rarer, I don't think we should complicate the %() support to make
  > it easy to define non-ansi macros.
Agreed.

  > One problem with this is that it means specs aren't backward compatible,
  > which means that people using -b or -V may get errors.  These options have
  > have never been safe, and I have always recommended against their use, but we
  > may need to keep them working.  I could perhaps just give a warning instead
  > of an error when %[] is used.
I'm not too concerned about this.

  > Thu Jul  9 15:24:18 1998  Jim Wilson  <wilson@cygnus.com>
  > 
  > 	* i386/i386.h (CPP_486_SPEC, CPP_586_SPEC, CPP_686_SPEC): New specs.
  > 	(CPP_CPU_DEFAULT_SPEC, CPP_CPU_SPEC): Use them.
  > 	(EXTRA_SPECS): Support them.
  > 	* gcc.c: Delete %[spec] support.
  > 	(do_spec_1, case '('): Likewise.
  > 	(do_spec_1, case '['): Call error.
  > 	* i386/aix386ng.h, cygwin32.h, freebsd-elf.h, gas.h, isc.h,
  > 	linux-aout.h, linux-oldld.h, linux.h, osfelf.h, osfrose.h, sco.h,
  > 	sco4.h, sco4dbx.h, sco5.h, sol2.h, sysv3.h (CPP_SPEC): Delete
  > 	%[cpp_cpu].
I went ahead and committed the patch since I'm trying to get the egcs-1.1
branch made tonight.

jeff