Re: PATCH: DG/UX Intel port

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

  In message <079b01be9bee$18bbb8b0$65c8c8c8@ehrpc.listworks.com>you write:
  > The first patch is required to build egcs at all.  When built this way, egc
  > s
  > will fail the stage2-to-stage3 compare operation done as part of "make
  > bootstrap".  This is because the DG/UX assembler inserts some time stamp
  > information into the object files which cause cmp to find differences.
  > However, these are only object file differences.  The executables compare
  > OK.  You'll need to continue the library builds after the make aborts. [Is
  > there a way to fix this?]
Several notes.  Please look at the format of all the existing ChangeLog entries
and try to follow that same format.

Please read the GNU coding standards and try to follow them.  For example, do
not format comments like this

/*
 * blah blah blah
 */

Instead:
/* blah blah blah
   more blah blah.  */


On a general note, instead of commenting out code, delete it.  If you feel
it necessary to leave a comment about why the code was deleted, that is fine.
But in general we frown on lots of #if 0 or commented out code.

Removing flag_signed_bitfields from OPTIMIZATION_OPTIONS is both good and bad.

The problem (I suspect) is dgux's compilers had unsigned bitfields and the
the in OPTIMIZATION_OPTIONS is trying to make gcc be compatible.  We need to
find a clean way to fix this problem.

For example we could have a macro DEFAULT_SIGNED_BITFIELDS

In c-decl.c (and other files which define & initialize flag_signed_bitfields)
we would do something like

#ifndef DEFAULT_SIGNED_BITFIELDS
#define DEFAULT_SIGNED_BITFIELDS 1
#endif

int flag_signed_bitfields = DEFAULT_SIGNED_BITFIELDS

Then dgux could define DEFAULT_SIGNED_BITFIELDS to zero to be compatible with
the dgux compilers.

In general, please don't gratutiously change the formatting of code unless you
are fixing a GNU coding standards violation.  When you change formatting it
just makes the patch & your code much more difficult to understand.

CPP_PREDEFINES:

  We are trying to get away from the -Di386 and -Dunix in CPP_PREDEFINES since
  those are violations of the ANSI namespace.  It would be good if you could
  convert dgux.h to CPP_SPEC to do this.  See linux.h  Similarly for ix86 and
  DGUX.
  
  I also don't think given -Di386 in cpp_spec that you'll need -D__i386__ or
  -D__i386.  I believe -Di386 will expand into all three as needed.  Similarly
  for -Dix86


Your patch adds random newlines.  Please don't do that.

You completely reformatted ASM_SPEC and make it more difficult to read rather
than less difficult.  Don't do that.  I can't even tell if you actually changed
anything of meaning without looking at your code _very_ closely.  Similarly
for your change to LIB_SPEC LINK_SPEC and STARTFILE_SPEC.


The change to t-dgux looks pretty reasonable.  I believe it merits a complete
and separate patch on its own so that we do not hold up a correct change while
you fix up your other changes.

Your changes to fixinc.dgux also included random, unnecessary whitespace
changes.  Don't do that.

I haven't looked closely at the second patch yet.

jeff