3.0.1 PATCH: install.texi updates for Solaris 2

[Rainer Orth]

Gerald Pfeifer writes:

[I'm copying the SPARC maintainers to comment on the status of SPARC V9
support on the 3.0 branch, as summarized in

	http://gcc.gnu.org/ml/gcc-patches/2001-08/msg00107.html

]

> On Thu, 2 Aug 2001, Rainer Orth wrote:
> > This patch updates the Solaris 2 parts of install.texi.
> >
> > Two questions about this stuff:
> >
> > * This section
> >
> > Unfortunately, C++ shared libraries, including @samp{libstdc++}, won't work
> > properly if assembled with Sun @command{as}: the linker will complain about
> > relocations in read-only sections, in the definition of virtual
> > tables.  Also, Sun @command{as} fails to process long symbols resulting from
> > mangling template-heavy C++ function names.
> >
> >   seems obsolete.  The introduction of the new ABI should have fixed part
> >   of this, and I never had any problems during my testing on Solaris 2.5.1
> >   and 8.  Can anyone confirm that this can go?
> 
> Yes. The issue with long symbols is definitely gone with the new mangling
> scheme (and we're using very deeply nested STL structures), and also the
> dynamic linking issue didn't appear recently.

ok, so I'll just remove this paragraph unless someone else objects.

> > * Could someone review the section on missing/incomplete 64-bit SPARC V9
> >   support to ensure it's both complete and accurate?  It documents my
> >   current understanding of the situation, but may well be wrong.
> 
> This needs someone else then me, unfortunately,...

I'm copying the SPARC maintainers for comments.  Since this issue comes up
over and over again both on gcc mailing lists, PRs and various Solaris
newsgroups, we should document that current state of affairs.  If my
assessment is accurate, it will go on the branch only, since all this is
supposed to be fixed on the trunk.

> > Tested with texi2dvi and makeinfo.
> >
> > Ok for branch and mainline?
> 
> ...but the rest is basically fine if it also passes
> gcc/doc/install.texi2html and we can resolve the Solaris 2 vs Solaris 7
> issue pointed out below.

Ok, I'll check this and post the final version, incorporating your
suggestions below, once feedback on 64-bit support has arrived.

> >  For example, the linker may hang indefinitely.  The fix is to remove
> >  @file{/usr/ucb} from your @code{PATH}.
>                              ^^^^^
> @env (and path in lower-case?)

@env is certainly right, but the variable is really called PATH (path is
just a special-case variable in csh-style shells, whose content is mirrored
to PATH).

> > +Starting with Solaris 7, the operating system is capable of executing
> 
> U-oh! If we're referring to Solaris 2 all the time, and want this term
> to subsume also Solaris 7, we have to call the latter Solaris 2.7.
> 
> Do you see what I mean? Respectively, do you have a better solution?

I do, but this would add to the current confusion: Solaris 2 releases
between 2.0 and 2.6 were called this, but Sun marketing changed 2.7 to 7
after the beta test cycle, i.e. very shortly before the release.  Solaris 8
and 9 have remained this way.  I don't want to call Solaris 7 by anything
but by its real name.  Basically, we need two things:

* a way to collectively refer to (a group of or all) Solaris 2 releases,
  i.e. 2.0 - 2.6, 7, 8, ...  I suggest to use Solaris 2 for this purpose,
  since this shouldn't include Solaris 1 (formerly SunOS 4.x) which is
  completely different.  This is what the Solaris 2 FAQ does, see

	http://www.wins.uva.nl/pub/solaris/solaris2/

* a way to refer to individual releases, like Solaris 2.5.1 or Solaris 7.
  Instead of using e.g. Solaris 2.7 here (which doesn't exist), I'd like to
  use the `real' name here.

I know this is terribly confusing, but some Sun marketing droid is guilty
of all this mess, and we shouldn't add to it ourselves.  So ok to leave the
Solaris 7 reference as is?

> >  partial fix is adequate for GCC@.  Revision -08 or later should fix
> > -the bug, but (as of 1999-10-06) it is still being tested.
> > +the bug.  The current revision is -13 (2000-10-02), and is included in
> > +the Solaris 7 Recommended Patch Cluster.
> 
> Please let's keep the "as of", for example "current (as of)" and use
> the current date.

Ok, will do.

	Rainer

-----------------------------------------------------------------------------
Rainer Orth, Faculty of Technology, Bielefeld University

Email: ro@TechFak.Uni-Bielefeld.DE