Re: Adding support for AS argument "-force-long-branchs" for x86arch

[Nick Clifton <nickc at redhat dot com>]

Hi Kristis,

> I had requested to the GCC developers the feature of  adding support
> for the -force-long-branchs argument in the AS assembler for the x86 
> architecture, as decsibed in
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12502. They have redirected
> me to the binutils maintainers.
>
> Please advice if adding this feature is possible. (Please read the
> entire bugzilla entry)

The feature is possible, although as far as I know nobody is working
upon adding it.  If you would like to work on this feature please feel
free.

The bugzilla case does expose a problem with the assembler however -
it assumes that the "-force-long-branches" switch is actually "-f",
meaning "do not remove pre-processor syntax from the input file",
which is why you are getting those strange error messages.

I am going to apply the patch below to fix this, so that now running:

  gcc -Wall -Wa,-force-long-branchs -O2 test10.c

will generate:

  as: unrecognised option `-force-long-branches'

Cheers
        Nick

gas/ChangeLog
2003-10-04  Nick Clifton  <nickc@redhat.com>

	* as.c (std_shortopts): Remove 'f'.
	(std_longopts): Add 'f'.  Doing this prevents -f<foo> being
	acecpted as an alias for -f.

Index: gas/as.c
===================================================================
RCS file: /cvs/src/src/gas/as.c,v
retrieving revision 1.46
diff -c -3 -p -r1.46 as.c
*** gas/as.c	4 Jun 2003 16:54:45 -0000	1.46
--- gas/as.c	4 Oct 2003 12:01:12 -0000
*************** parse_args (pargc, pargv)
*** 375,381 ****
      /* -K is not meaningful if .word is not being hacked.  */
      'K',
  #endif
!     'L', 'M', 'R', 'W', 'Z', 'f', 'a', ':', ':', 'D', 'I', ':', 'o', ':',
  #ifndef VMS
      /* -v takes an argument on VMS, so we don't make it a generic
         option.  */
--- 375,381 ----
      /* -K is not meaningful if .word is not being hacked.  */
      'K',
  #endif
!     'L', 'M', 'R', 'W', 'Z', 'a', ':', ':', 'D', 'I', ':', 'o', ':',
  #ifndef VMS
      /* -v takes an argument on VMS, so we don't make it a generic
         option.  */
*************** parse_args (pargc, pargv)
*** 449,454 ****
--- 449,460 ----
  #define OPTION_NOEXECSTACK (OPTION_STD_BASE + 21)
      {"noexecstack", no_argument, NULL, OPTION_NOEXECSTACK},
  #endif
+     /* Treat '-f' as a long switch so that getopt will not accept
+        -f<some-text> as a synonym for -f.  This can cause confusion
+        when -f switches are passed through from the compiler.
+        FIXME - should we handle other single character switches in the
+        same way ?  */
+     {"f", no_argument, NULL, 'f'},
  #define OPTION_WARN_FATAL (OPTION_STD_BASE + 22)
      {"fatal-warnings", no_argument, NULL, OPTION_WARN_FATAL}
      /* When you add options here, check that they do not collide with