Remove lang/gcc32 (GCC 3.2) which is no longer used by any other ports and
does not support amd64, for example.

- Remove support for a.out format and PORTOBJFORMAT variable from individual
  ports

With hat:       portmgr

- fix build with gcc41
Obtained from:  upstream

- take benefit of USE_LDCONFIG

- take benefit of USE_LDCONFIG

Remove USE_REINPLACE from the categories starting with L

- SUFFIX is not defined here

- add RC_D_SH to keep shared libs working after reboot

Revert last commit.  It is not possible (yet) to have the dirrmtry usage
in the post-install target.  Its usage must preceed generate-plist.

This problem needs to be revisited either in a rework of bsd.port.mk or
each of these Makefiles.

Hat:            portmgr
Pointy hat:     edwin`

eplace ugly "@unexec rmdir %D... 2>/dev/null || true" with @dirrmtry

Approved by:    krion@
PR:             ports/88711 (related)

Remove install-info from Makefile, it's automatically done when INFO is defined

Use INFO macro

- Add SHA256

Sync freebsd7 support with gcc33.

Sync freebsd7 support with gcc33.

Fix build on 7-CURRENT

(forced commit)

Submitted by: "Jack L." <xxjack12xx@doramail.com>

Define WITHOUT_CPU_FLAGS so that in some cases
setting CPUTYPE breaks the build

Remove from the Java category as this port doesn't have Java bits anymore.
Also remove the testsuite bits as any issues found will never be fixed.

It was discovered that running "make clean" from a port with a dependency
on gcc31 or gcc32 (either with a BUILD_DEPENDS or via USE_GCC) with that
version of gcc already installed causes infinite recursion problems.

The issue stems from gcc31 and gcc32 using USE_GCC internally. Internally,
it was used to override any other definition and say that you can use any
gcc version to build gcc31/gcc32. But that USE_GCC redefinition was causing
recursion errors, so handle the problem instead by just simply undefining
USE_GCC inside the gcc port Makefile itself.

This should make things happy again.

Fix USE_GCC

PR:             ports/77403
Submitted by:   John Cochran

This now coredumps when bootstrapping on FreeBSD/AMD64.

o Fix -pthread issue.
  This fix resolves openoffice building problem on 5.2.1 or later.
o Bump PORTREVISION.

PR:                     ports/75785 [1], bin/62300 [2]
Pointed out by:         Oleg Sharoiko <os@rsu.ru> [1],
                        many many people on openoffice@FreeBSD.org
Inspired by:            maho, Mark Andrews <marka@isc.org> [2]
Discussed with:         maho, daichi
Committed at:           Murayakuba Japanese style pub
                        http://www.murayakuba.jp/
Committed with:         maho, daichi

Cleaner support -pthreads for 5.0-5.2.1.

Fix xfmode splits on AMD64; using patch from Richard Henderson <rth@redhat.com>

Unbreak the port by remove the Java language from this port.  Anyone
using GCC 3.2 isn't using it for the Java compiler, but for C/C++ code.

Adjust spec to make -pthread and gcj link to -lpthread on recent
FreeBSD 5.

- drop maintainership

- mark BROKEN on amd64

- fix configure for amd64

- mark BROKEN on ia64

Bootstrap comparison failure!
java/decl.o differs

- fix typo in condition.
Reported by:    kris

- fix amd64 build
- add option WITHOUT_LIBJAVA=yes

- take maintainership
gcc33 objc compilers don't compile gnustep-base.
so this port is still needed for a bit.

- let build also on alpha and sparc64
bento succesfull build logs:
i386-5-latest-logs      2004 Apr 11 09:50:21    gcc-objc-3.2.3_1.log
ia64-5-latest-logs      2004 Apr  8 18:37:05    gcc-objc-3.2.3_1.log
alpha-5-latest-logs     2004 Apr  4 12:34:15    gcc-objc-3.2.3_1.log.bz2
alpha-4-latest-logs     2004 Mar 25 20:11:44    gcc-objc-3.2.3_1.log.bz2
sparc64-5-latest-logs   2004 Mar 22 15:23:03    gcc-objc-3.2.3_1.log.bz2

This old version of GCC really only makes sense on i386 at this point (if at
all), so use ONLY_FOR_ARCHS=i386.  Drop maintainership.

Remove WITH_BOUNDSCHECKING knob, as the patch disappeared upstream and
GCC 3.3 with a current bounds checking patch should be better anyways.

Reported by:    trevor

Mark broken on amd64 (which used to be called x86_64).

Bump PORTREVISION on all ports that depend on gettext to aid with upgrading.

(Part 2)

Use the SORT macro from bsd.port.mk.

Fix the incorrect pkg-descr on binaries installed and "sign" this port.
Given that the GCC 3.2 branch has been completely frozen, remove CVS and
update-related features from Makefile.

Make portlint(1) happy by changing strip to ${STRIP_CMD}

Submitted by:   Oleg Karachevtsev <ok@etrust.ru>
PR:             56998

- remove empty dir lib/security

Track upstream update of the patch for WITH_BOUNDSCHECKING (which removed
the previous patch).

Submitted by:   lioux

Remove three obsolete comments.

Remove --with-gnu-as --with-gnu-ld, which are actually no-ops.
Also, we don't need USE_BISON for GCC releases.

Fix install problem of gnustep-objc
extend comment for ia64

There is now a version of the bounds-checking patch also for GCC 3.2.3.

Submitted by:   Dan Nelson <dnelson@allantgroup.com>

Merge targets pre-configure and pre-build, simplify them, and remove
patch-va.  All of these dealt with us patching generated files (which
we don't do any longer).

Update to GCC 3.2.3 as released today.  (This is the last release from this
branch which has been "closed" in GCC CVS, so we can simplify this port.)

Make MAINTAINER overrideable by slave ports.

- drop javlib in PLIST for ia64 too.
Approved by:    gerald

On sparc64 libgcj is not built by default (and won't build easily), so we
need to dynamically construct pkg-plist to take this into account.

Noticed by:     bento+kris

Remove the last part of jdp's profile fixes from 1998/10/08, which actually
breaks -fprofile-arcs for current GCCs.

PR:             50842

1) COMMENT can be overrritten by a slave port (suggested by obrien)
2) Option WANT_SHAREDLIBS now uses ldconfig, so LIB_DEPENDS work.
   fix PLIST so critical file libgcc_s.so.1 is included.
   no response from obrien since 2003-03-31
Approved by:    gerald

Remove PATCH_SITES and PATCHFILES which we haven't used lately and related
material.

Reviewed by:    obrien

Appease portlint by using ${GMAKE}, ${GREP}, and some whitespace changes.

Finally remove the parts of this patch that date back to 1998 and have now
been commented for more than a year.

Simplify the shell code used by the pre-patch target.

Suggested by:   obrien
Reviewed by:    Dan Nelson <dnelson@allantgroup.com>

LATEST_LINK can now be overwritten by the slave port.
added ${PKGNAMESUFFIX} to LATEST_LINK, so it differs when set.
Approved by:    gerald

Add support for the bounds-checking patch by Herman ten Brugge (which
provides an -fbounds-checking command-line option).  This is optional
and built only if WITH_BOUNDSCHECKING is defined.

PR:             50178
Submitted by:   Dan Nelson <dnelson@allantgroup.com>

Add support for the base language test suites.

Submitted by:   Loren James Rittle <rittle@labs.mot.com>

Enable libgcj (which has actually been building fine since GCC 3.1 on
FreeBSD/i386 and works on FreeBSD/alpha using the hack in files/patch-ub).

PR:                     47790
Partially Submitted by: Loren James Rittle <rittle@labs.mot.com>

Remove special alpha optimization warning.  alpha support now should be
sufficiently stable, and the testsuite output for FreeBSD/alpha shows no
extra -O2 failures.

Submitted by:   Loren James Rittle <rittle@labs.mot.com>

Assume maintainership of this port.

Approved by:    obrien (mentor)

Fix COMMENT and a Makefile comment.

Remove special USER_H override.

Submitted by:   Loren James Rittle <rittle@labs.mot.com>
Approved by:    kris (portmgr)

De-pkg-comment.

Bump PORTREVISION to the 5-Feb-2003 release date of GCC 3.2.2 to make sure
all the various tools knows a newer version is available than anything
installed to date.

Update to GCC 3.2.2.

Approved by:    obrien (mentor)

Adjust pkg-comment, this is not 3.1.

PR:             44374
Submitted by:   Volker Stolz <vs@foldr.org>

Use the most modern spelling of egcs.cygnus.com.

Fix cklatest paths.

Reported by:    gerald
Approved by:    kris

Update to 3.2.1 release, and fix packaging.

The effort to maintain these for FreeBSD just isn't worth it with the lack
of reward.

Use MASTER_SITES_GCC.

Approved by:    obrien (maintainer)

Add CVS "fetch" framework.

Total hack to allow gcc32 to build on -CURRENT.

This hack makes gcc totally use our stock headers vs. the fixinc'ed ones.
I don't like this change as I don't think there is any chance of getting
this change back into the GCC repository and it is important that the
stock GCC sources will bootstrap on FreeBSD 5.0.

Submitted by:   Don Lewis <dl-freebsd@catspoiler.org>

Upgrade to the 2-Sept-2002 GCC 3.2.1 pre-release snapshot.

(this port has been 100% reset back to an earlier evolution line by a repo
copied from the 'gcc31' port)

ftp://ftp.netwinder.org/ is no longer a MASTER_SITE.

Approved by:    obrien (maintainer+mentor)

This is now in the stock FSF sources.

Upgrade to a 18-May-2002 GCC 3.2 in-development checkout.

GCC 2.95 and later aren't called egcs (any longer).  Also update
description and web address.

Approved by:    obrien (maintainer+mentor)

GCC 2.95 and later aren't called egcs (any longer), so remove two patches
that add references to egcs to the documentation.

Approved by:    obrien (maintainer+mentor)

Upgrade to a 06-April-2002 GCC 3.2 in-development checkout.

Upgrade to a 27-March-2002 GCC 3.2 in-development checkout.

Upgrade to a 12-March-2002 GCC 3.2 in-development checkout.

Upgrade to the 10-March-2002 GCC 3.2 in-development snapshot.

Due to the lack of a real snapshot distfile, this has to use CVS
as the fetch method.

Mark broken while I find a suitable snapshot.