Re: OpenBSD configuration for main platforms, framework

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

  In message <19990216184117.A20285@tetto.liafa.jussieu.fr>you write:
  > 
  > --5vNYLRcllDrimb99
  > Content-Type: text/plain; charset=us-ascii
  > 
  > This is the first patch of what's needed to configure egcs on the most
  > critical openbsd platforms (i386, m68k, sparc, and alpha), 
  > plus a config.guess patch so openbsd-*-hppa gets at least recognized...
Thanks.


  > Feel free to tweak the comments to reflect FSF copyright and remove
  > the $OpenBSD$ tags... I assume you've got some automatic mechanism to do 
  > that, and I don't want to attach n copies of this.
Please submit the patches with fixed comments and removed the RCS stuff.  When
you don't do this, I have to, for each and every one of your files.

  > ChangeLog for gcc:
  > 
  > Tue Feb 16 18:30:13 CET 1999	Marc Espie <espie@cvs.openbsd.org>
  > 	* configure.in : configure openbsd platforms
  > 	* config.guess: recognize openbsd-*-hppa
  > 	* config/openbsd.h: New file
  > 	* config/xm-openbsd.h: New file
  > 	* config/t-openbsd: New file
  > 	* config/t-openbsd-thread: New file, experimental thread support
Capitolize and periods appy to ChangeLog entries too.


  > /* $OpenBSD: openbsd.h,v 1.13 1999/02/16 17:20:55 espie Exp $	*/
Kill.

  >    There are also a few `default' defines such as ASM_WEAKEN_LABEL,
  >    intended as common ground for arch that don't provide anything suitable.
  >  */
Close the comment on the previous line.  If it won't fit, put the last word
in the comment on a new line, then close the comment.

  > 
  > /* OPENBSD_NATIVE is defined only when gcc is configured as part of
  >    the OpenBSD source tree, specifically through Makefile.bsd-wrapper.
  > 
  >    In such a case the include path can be trimmed as there is no
  >    distinction between system includes and gcc includes
  >  */
  > /* This configuration method, namely Makefile.bsd-wrapper and
  >    OPENBSD_NATIVE is NOT recommended for building cross-compilers
  >  */
Likewise.

  > 
  > /* Controlling the compilation driver 
  >    ---------------------------------- */
Kill ------------

  > /* CPP_SPEC appropriate for OpenBSD. We deal with -posix and -pthread.
  >    XXX the way threads are handling currently is not very satisfying,
  >    since all code must be compiled with -pthread to work. 
  > 	This two-stage defines makes it easy to pick that for targets that
  > 	have subspecs.
  >  */
Comment format.

  > 
  > #ifndef OBSD_NO_DYNAMIC_LIBRARIES
  > #undef SWITCH_TAKES_ARG
  > #define SWITCH_TAKES_ARG(CHAR) \
  >   (DEFAULT_SWITCH_TAKES_ARG(CHAR) \
  >    || (CHAR) == 'R')
  > #endif
Space before the open paren for the args to DEFAULT_SWITCH_TAKES_ARG.

  > /* Runtime target specification 
  >    ---------------------------- */
Fix.

  > /* Implicit calls to library routines
  >    ---------------------------------- */
Fix.

  > /* Miscellaneous parameters
  >    ------------------------ */
  > /* tell libgcc2.c that OpenBSD targets support atexit. */
Fix.


  > /* Controlling debugging info: dbx options 
  >    --------------------------------------- */
  > /* Don't use the `xsTAG;' construct in DBX output; OpenBSD systems that
  >    use DBX don't support it. */
Fix.

  > /* Support of shared libraries, mostly imported from svr4.h through netbsd.
  >  */
Fix.


  > /* Assembler format: output and generation of labels
  >    ------------------------------------------------- */
Fix.

  > /* Storage layout 
  >    -------------- */
Fix.

  > /* We don't have to worry about binary compatibility with older C++ code,
  >    but there is a big known bug with vtable thunks which has not been
  > 	fixed yet, so DON'T activate it by default.
  >  */
Fix.


  > /* Otherwise, since we support weak, gthr.h erroneously tries to use
  >    #pragma weak
  >  */
Fix.


  > /* `generic' xm-openbsd.h.
  >    This file gets included by all architectures. It holds stuff
  > 	that ought to be defined when hosting a compiler on an OpenBSD
  > 	machine, independently of the architecture. It's included by
  > 	${cpu_type}/xm-openbsd.h, not included directly.
  >  */
Fix.


  > /* OpenBSD is trying to be POSIX-compliant, to the point of fixing
  >    problems that may occur with gcc's interpretation.
  >  */
Fix.



Overall the patches look pretty good and probably can go in once you get
the copyright headers fixed and the formatting stuff.  I guess I won't
scream too much more about the OPENBSD_NATIVE stuff, even though I think
you're still making a long term mistake.  A better way to do that is to
use configure options to control more closely what directories are used.

Thanks,

jeff