riscv64 is a LP64_ARCH, list it as such.

Retire OpenBSD/sgi.

OK deraadt@

Update compiler-rt path in cross-build infrastructure.

Initial set of changes to build powerpc64 toolchain.

Disable PIC/PIE for powerpc64 for now.

Simplify by getting rid of the unecessary 'depend' rule.

ok patrick@

Add mips64el to LP64_ARCH.

OK patrick@

Re-sort the if-sequence which sets COMPILER_VERSION to be the same
as bsd.own.mk, otherwise on transitioning architectures the wrong
compiler is chosen. Also make sure BUILD_CLANG is forced for the
libcompiler_rt directory since cross-compiling with two possible
compilers would create too much headache.

Fix whitespace issues

powerpc cross-compilation on 64-bit hosts also appears to be safe.

Pass LINKER_VERSION to binutils and clang cross-build so that they
know who the system linker is.

Fix malformed conditional after socppc removal.

From Krystian Lewandowski

socppc makes an extended visit to the bigbucket.
ok kettenis

Don't assume that we use LLVM's ld(1) if we use clang(1) as main system
compiler. Only handle this case when LLD was actually installed during
the cross-compiler setup. Unbreaks armv7 after switching to clang since
it uses binutils' ld(1).

Compile the cross-compiler in the target's objdir instead of polluting
the host's. This also allows compiling multiple toolchains in parallel.
Clean up the clang object files afterwards so it doesn't collide with
the cross-compiled compiler build.

Fix cross-compiling GCC toolchains on Clang archs.

ok patrick@

Fix cross builds: no clang depend target, no DESTDIR on HOSTCC build
ok patrick@

Pass CC and CXX to make depend as well. As soon as we support
dependencies in the clang makefiles we need to make sure to use
a compiler that supports C++11.

Adjust cross-toolchain environment to the changes in the clang build
fabric. Since we now also build and use LLVM's lld we can copy lld to
the proper directory akin to what we do with clang. Build compiler-rt
early as it is needed for building libc.

"go for it" mpi@

Export CROSSDIR so that the two users (ld.so and binutils) of that
variable actually work correctly. While there, adjust a bit for
consistency.

"nobody will complain if you break it ;)" kettenis@

We can now build parts of binutils on aarch64 so enable cross-binutils
again.

ok patrick@

Adjust cross-compilation infrastructure to cope with builduser
permission requirements.

From Dale Rahn.

ok tb@
"commit it" guenther@

With clang as cross-compiler we need to tell it where our cross
directory is so that the correct include and library paths are used.
For this we create a simple wrapper that calls clang in the correct
mode and also passes the sysroot.

ok kettenis@

Implement support for cross-compiling arm64 in Makefile.cross. This
adds code to build clang as cross-compiler instead of gcc. Since we
don't have binutils for aarch64 in base, skip the binutils cross-
toolchain build for that architecture.

ok kettenis@

Unhook sqlite3.

Retire zaurus, as it hasn't made the EABI jump and will be permanently broken
shortly when we use the hardware thread register in userland

armish handled some early-gen arm machines, which required tons of
workarounds. Some of them will soon stand in the way of armv7.
Off to the attic you go.

remove hppa64 port, which we never got going beyond broken single users.
hppa reverse-stack gives us a valuable test case, but most developers don't
have a 2nd one to proceed further with this.
ok kettenis

We are done providing support for the vax.
lots of agreement.

Restore parenthesis

Send solbourne to where the ...........

Send Aviion to same place as Nova II
discussed with jsg

libocurses can go to the Attic. last consumer of it was ramdisk more(1)
conceptual ok guenther millert nicm

Always honor ${DESTDIR}. OK miod@

Use target BSDOBJDIR when cross-compiling libraries

ok mpi@

Now that we do not need to build binutils 2.15 (except for gdb), simplify the
build logic.

Invert the binutils 2.17 logic to only list non-2.17 platforms.

Fix cross-build. OK jsing@ miod@ guenther@ deraadt@

Remove the last gzsig tendril

more NO_CROSS

lynx will go to ports. there are too many reasons for it to go there,
and not enough for it to stay. lengthy discussions.

Make sure to override PIE_DEFAULT when building a cross-gcc4

Remove more Kerberos cruft.

ok jca@ jmc@

Retire hp300, mvme68k and mvme88k ports. These ports have no users, keeping
this hardware alive is becoming increasingly difficult, and I should heed the
message sent by the three disks which have died on me over the last few days.

Noone sane will mourn these ports anyway. So long, and thanks for the fish.

Beagle has been replaced by armv7.

Remove support for COMPILER_VERSION == gcc2.
Change the logic depending upon COMPILER_VERSION everywhere, to assume gcc4
is the norm and to explicitely test for gcc3 when a different behaviour
is required.
No functional change intended. Be sure to `make install' in share/mk before
attempting to do anything.

tabs got lost in previous rev

ELF_TOOLCHAIN bye bye.

remove BIN{MODE,GRP,OWN} since bsd.own.mk is already included.

ok bmercer@

gumstix was never finished

the unused palm code goes away.

Allow the PIE/NOPIE variables in bsd.own.mk to be overriden. Force proper
values in Makefile.cross to make sure binutils are built with the proper
PIE_DEFAULT value for the target.
ok millert@ deraadt@

m68k ports are now ELF citizens.

Switch m88k ports to ELF.

Pass -DOPENBSD_CROSS to CFLAGS when cross-compiling gcc2. Temporary bandaid
until we get rid of it. Soon.

Make COMPILER_VERSION a definitive assignment in there, for it to override
<bsd.own.mk>. Allows cross-gcc to be built on a host using a different compiler
version than the target.

no need to know about afs, since it is gone

Garbage collect the USE_GCC variables; everything uses
COMPILER_VERSION now.

From Brad.

remove leftover NOLINT, WANTLINT, LINTFLAGS, LOBJ vars and lint targets.

ok guenther@

Add sqlite3 to NO_CROSS since it depends on libreadline - this would then
require cross-gnu to build before we can build cross-bin.

more mac68k bits for the attic

Do not try to cross build nginx, since it does not work.

- add avr32 which'll use gcc4/binutils-2.17
- remove BINUTILS217_ARCH from Makefile.cross as bsd.own.mk is included there
(requested by drahn@)

ok drahn@ deraadt@

Allow 64bit -> arm cross compilation as it appears safe (amd64 -> arm at least)
include bsd.own.mk to get the correct compiler architectures
(if only Makefile.cross is used)

Detect word-size mismatches between host and target, which result in
the cross compiler generating bad code, and abort

ok deraadt@

ia64 should use gcc4; sort the list of binutils-2.17 platforms while I'm there

ok jsing@, miod@

Switch hppa64 cross compiling environment to gcc4.

Switch hppa64 to binutils-2.17

remove mvmeppc; it is really rough shape. ok drahn miod

Allow switching between binutils and binutils 2.17.
NOTE: binutils-2.17 has not been tested on any archs. ok pirofti@

Typo, BINOWN -> BINGRP

ok miod@

unlink libsectok from the build; ok markus@ deraadt@

Copy the right bits to the right places for gcc4.

Switch mips-based ports to gcc4. Most of the work has been done by kettenis@.

Remove tn3270 reference.

add octeon as cross build target arch. ok miod@

nsd uses configure, so put it in NO_CROSS for now.

Sigh, check in the actual diff, not a backout of th previous change.
Use TARGET_ARCH, not MACHINE_ARCH. Fix from Tobias Ulmer. ok jsg@ jasper@

Use TARGET_ARCH, not MACHINE_ARCH. Fix from Tobias Ulmer. ok jsg@ jasper@

sync the list of gcc4 archs with bsd.own.mk
ok miod@

Fix cross build problem with cpp, marex pointed out that it wasn't committed.

Initial attempt at being COMPILER_VERSION and gcc4 compatible.

moko goes away.

Force MACHINE_CPU when building cross-toolchain or cross-building. This allows
loongson cross-tools to be configured correctly.

Add all arm based architectures. ok miod@.

Palm: Integration into the tree

OK deraadt@

Try to clean up the cross build environment, more of the tree cross builds now
ok miod@

socppc is powerpc

another cats remnant

feed bsdsrc/objdir and makeobjdir in cross-env allowing make obj and the like to work on random directories; found w/ rainerx0r; drahn@ ok

missing TARGET_ARCH handling (solbourne)

Early bits for a 32-bit SuperH toolchain, currently for little-endian
systems only; more work is necessary in ld land.

alpha is gcc3 now

typoo

System build pieces for armish arch.

Administrative changes for aviion, ok deraadt@

Install binutils objdump on a.out cross-tools; ok grange@

make sure cross-as and cross-ranlib get compiled again for a.out platforms.

photurisd is nomore; from Michael Knudsen

Disable lint for libraries.

ok deraadt@ espie@

Fix cross-etc-root-var, pick up target etc files, simplied fix from PR4511
ok grange@

Compute TARGET_ARCH in the Makefile, rather than in a shell, and use it
instead of MACHINE_ARCH to pick compiler and binutils versions.

looked at mickey@ grange@

Create proper symlink for a.out as.

Create a correct /usr/bin/cpp for a.out platforms.

telnetd is gone

Kill \n at EOF.

match gcc3 logic from bsd.own.mk and add a target_arch generation for arm

No need for cross nm/size in cross-env, our host tools are good enough.

ok mickey@

Unbreak cross-tools by providing TARGET_ARCH in include target
to get right config.status file for binutils and gcc.

ok mickey@

less confusing `TARGET_USE_GCC3' since it's only used locally.
Okay niklas@

- introduce TARGET_CANON/TARGET_ARCH shell variables, to make things slightly
clearer.

- don't put full path into symlinks that don't change directory.

- add a few NO_CROSS directories: stuff that depends on libcrypto, stuff
that needs host tools, and stuff that wants cross-configure.

okay niklas@

move cross-stuff into its own file.
okay mickey@, niklas@

Retire OpenBSD/sgi.

OK deraadt@

Update compiler-rt path in cross-build infrastructure.

Initial set of changes to build powerpc64 toolchain.

Disable PIC/PIE for powerpc64 for now.

Simplify by getting rid of the unecessary 'depend' rule.

ok patrick@

Add mips64el to LP64_ARCH.

OK patrick@

Re-sort the if-sequence which sets COMPILER_VERSION to be the same
as bsd.own.mk, otherwise on transitioning architectures the wrong
compiler is chosen. Also make sure BUILD_CLANG is forced for the
libcompiler_rt directory since cross-compiling with two possible
compilers would create too much headache.

Fix whitespace issues

powerpc cross-compilation on 64-bit hosts also appears to be safe.

Pass LINKER_VERSION to binutils and clang cross-build so that they
know who the system linker is.

Fix malformed conditional after socppc removal.

From Krystian Lewandowski

socppc makes an extended visit to the bigbucket.
ok kettenis

Don't assume that we use LLVM's ld(1) if we use clang(1) as main system
compiler. Only handle this case when LLD was actually installed during
the cross-compiler setup. Unbreaks armv7 after switching to clang since
it uses binutils' ld(1).

Compile the cross-compiler in the target's objdir instead of polluting
the host's. This also allows compiling multiple toolchains in parallel.
Clean up the clang object files afterwards so it doesn't collide with
the cross-compiled compiler build.

Fix cross-compiling GCC toolchains on Clang archs.

ok patrick@

Fix cross builds: no clang depend target, no DESTDIR on HOSTCC build
ok patrick@

Pass CC and CXX to make depend as well. As soon as we support
dependencies in the clang makefiles we need to make sure to use
a compiler that supports C++11.

Adjust cross-toolchain environment to the changes in the clang build
fabric. Since we now also build and use LLVM's lld we can copy lld to
the proper directory akin to what we do with clang. Build compiler-rt
early as it is needed for building libc.

"go for it" mpi@

Export CROSSDIR so that the two users (ld.so and binutils) of that
variable actually work correctly. While there, adjust a bit for
consistency.

"nobody will complain if you break it ;)" kettenis@

We can now build parts of binutils on aarch64 so enable cross-binutils
again.

ok patrick@

Adjust cross-compilation infrastructure to cope with builduser
permission requirements.

From Dale Rahn.

ok tb@
"commit it" guenther@

With clang as cross-compiler we need to tell it where our cross
directory is so that the correct include and library paths are used.
For this we create a simple wrapper that calls clang in the correct
mode and also passes the sysroot.

ok kettenis@

Implement support for cross-compiling arm64 in Makefile.cross. This
adds code to build clang as cross-compiler instead of gcc. Since we
don't have binutils for aarch64 in base, skip the binutils cross-
toolchain build for that architecture.

ok kettenis@

Unhook sqlite3.

Retire zaurus, as it hasn't made the EABI jump and will be permanently broken
shortly when we use the hardware thread register in userland

armish handled some early-gen arm machines, which required tons of
workarounds. Some of them will soon stand in the way of armv7.
Off to the attic you go.

remove hppa64 port, which we never got going beyond broken single users.
hppa reverse-stack gives us a valuable test case, but most developers don't
have a 2nd one to proceed further with this.
ok kettenis

We are done providing support for the vax.
lots of agreement.

Restore parenthesis

Send solbourne to where the ...........

Send Aviion to same place as Nova II
discussed with jsg

libocurses can go to the Attic. last consumer of it was ramdisk more(1)
conceptual ok guenther millert nicm

Always honor ${DESTDIR}. OK miod@

Use target BSDOBJDIR when cross-compiling libraries

ok mpi@

Now that we do not need to build binutils 2.15 (except for gdb), simplify the
build logic.

Invert the binutils 2.17 logic to only list non-2.17 platforms.

Fix cross-build. OK jsing@ miod@ guenther@ deraadt@

Remove the last gzsig tendril

more NO_CROSS

lynx will go to ports. there are too many reasons for it to go there,
and not enough for it to stay. lengthy discussions.

Make sure to override PIE_DEFAULT when building a cross-gcc4

Remove more Kerberos cruft.

ok jca@ jmc@

Retire hp300, mvme68k and mvme88k ports. These ports have no users, keeping
this hardware alive is becoming increasingly difficult, and I should heed the
message sent by the three disks which have died on me over the last few days.

Noone sane will mourn these ports anyway. So long, and thanks for the fish.

Beagle has been replaced by armv7.

Remove support for COMPILER_VERSION == gcc2.
Change the logic depending upon COMPILER_VERSION everywhere, to assume gcc4
is the norm and to explicitely test for gcc3 when a different behaviour
is required.
No functional change intended. Be sure to `make install' in share/mk before
attempting to do anything.

tabs got lost in previous rev

ELF_TOOLCHAIN bye bye.

remove BIN{MODE,GRP,OWN} since bsd.own.mk is already included.

ok bmercer@

gumstix was never finished

the unused palm code goes away.

Allow the PIE/NOPIE variables in bsd.own.mk to be overriden. Force proper
values in Makefile.cross to make sure binutils are built with the proper
PIE_DEFAULT value for the target.
ok millert@ deraadt@

m68k ports are now ELF citizens.

Switch m88k ports to ELF.

Pass -DOPENBSD_CROSS to CFLAGS when cross-compiling gcc2. Temporary bandaid
until we get rid of it. Soon.

Make COMPILER_VERSION a definitive assignment in there, for it to override
<bsd.own.mk>. Allows cross-gcc to be built on a host using a different compiler
version than the target.

no need to know about afs, since it is gone

Garbage collect the USE_GCC variables; everything uses
COMPILER_VERSION now.

From Brad.

remove leftover NOLINT, WANTLINT, LINTFLAGS, LOBJ vars and lint targets.

ok guenther@

Add sqlite3 to NO_CROSS since it depends on libreadline - this would then
require cross-gnu to build before we can build cross-bin.

more mac68k bits for the attic

Do not try to cross build nginx, since it does not work.

- add avr32 which'll use gcc4/binutils-2.17
- remove BINUTILS217_ARCH from Makefile.cross as bsd.own.mk is included there
(requested by drahn@)

ok drahn@ deraadt@

Allow 64bit -> arm cross compilation as it appears safe (amd64 -> arm at least)
include bsd.own.mk to get the correct compiler architectures
(if only Makefile.cross is used)

Detect word-size mismatches between host and target, which result in
the cross compiler generating bad code, and abort

ok deraadt@

ia64 should use gcc4; sort the list of binutils-2.17 platforms while I'm there

ok jsing@, miod@

Switch hppa64 cross compiling environment to gcc4.

Switch hppa64 to binutils-2.17

remove mvmeppc; it is really rough shape. ok drahn miod

Allow switching between binutils and binutils 2.17.
NOTE: binutils-2.17 has not been tested on any archs. ok pirofti@

Typo, BINOWN -> BINGRP

ok miod@

unlink libsectok from the build; ok markus@ deraadt@

Copy the right bits to the right places for gcc4.

Switch mips-based ports to gcc4. Most of the work has been done by kettenis@.

Remove tn3270 reference.

add octeon as cross build target arch. ok miod@

nsd uses configure, so put it in NO_CROSS for now.

Sigh, check in the actual diff, not a backout of th previous change.
Use TARGET_ARCH, not MACHINE_ARCH. Fix from Tobias Ulmer. ok jsg@ jasper@

Use TARGET_ARCH, not MACHINE_ARCH. Fix from Tobias Ulmer. ok jsg@ jasper@

sync the list of gcc4 archs with bsd.own.mk
ok miod@

Fix cross build problem with cpp, marex pointed out that it wasn't committed.

Initial attempt at being COMPILER_VERSION and gcc4 compatible.

moko goes away.

Force MACHINE_CPU when building cross-toolchain or cross-building. This allows
loongson cross-tools to be configured correctly.

Add all arm based architectures. ok miod@.

Palm: Integration into the tree

OK deraadt@

Try to clean up the cross build environment, more of the tree cross builds now
ok miod@

socppc is powerpc

another cats remnant

feed bsdsrc/objdir and makeobjdir in cross-env allowing make obj and the like to work on random directories; found w/ rainerx0r; drahn@ ok

missing TARGET_ARCH handling (solbourne)

Early bits for a 32-bit SuperH toolchain, currently for little-endian
systems only; more work is necessary in ld land.

alpha is gcc3 now

typoo

System build pieces for armish arch.

Administrative changes for aviion, ok deraadt@

Install binutils objdump on a.out cross-tools; ok grange@

make sure cross-as and cross-ranlib get compiled again for a.out platforms.

photurisd is nomore; from Michael Knudsen

Disable lint for libraries.

ok deraadt@ espie@

Fix cross-etc-root-var, pick up target etc files, simplied fix from PR4511
ok grange@

Compute TARGET_ARCH in the Makefile, rather than in a shell, and use it
instead of MACHINE_ARCH to pick compiler and binutils versions.

looked at mickey@ grange@

Create proper symlink for a.out as.

Create a correct /usr/bin/cpp for a.out platforms.

telnetd is gone

Kill \n at EOF.

match gcc3 logic from bsd.own.mk and add a target_arch generation for arm

No need for cross nm/size in cross-env, our host tools are good enough.

ok mickey@

Unbreak cross-tools by providing TARGET_ARCH in include target
to get right config.status file for binutils and gcc.

ok mickey@

less confusing `TARGET_USE_GCC3' since it's only used locally.
Okay niklas@

- introduce TARGET_CANON/TARGET_ARCH shell variables, to make things slightly
clearer.

- don't put full path into symlinks that don't change directory.

- add a few NO_CROSS directories: stuff that depends on libcrypto, stuff
that needs host tools, and stuff that wants cross-configure.

okay niklas@

move cross-stuff into its own file.
okay mickey@, niklas@

Update compiler-rt path in cross-build infrastructure.

Initial set of changes to build powerpc64 toolchain.

Disable PIC/PIE for powerpc64 for now.

Simplify by getting rid of the unecessary 'depend' rule.

ok patrick@

Add mips64el to LP64_ARCH.

OK patrick@

Re-sort the if-sequence which sets COMPILER_VERSION to be the same
as bsd.own.mk, otherwise on transitioning architectures the wrong
compiler is chosen. Also make sure BUILD_CLANG is forced for the
libcompiler_rt directory since cross-compiling with two possible
compilers would create too much headache.

Fix whitespace issues

powerpc cross-compilation on 64-bit hosts also appears to be safe.

Pass LINKER_VERSION to binutils and clang cross-build so that they
know who the system linker is.

Fix malformed conditional after socppc removal.

From Krystian Lewandowski

socppc makes an extended visit to the bigbucket.
ok kettenis

Don't assume that we use LLVM's ld(1) if we use clang(1) as main system
compiler. Only handle this case when LLD was actually installed during
the cross-compiler setup. Unbreaks armv7 after switching to clang since
it uses binutils' ld(1).

Compile the cross-compiler in the target's objdir instead of polluting
the host's. This also allows compiling multiple toolchains in parallel.
Clean up the clang object files afterwards so it doesn't collide with
the cross-compiled compiler build.

Fix cross-compiling GCC toolchains on Clang archs.

ok patrick@

Fix cross builds: no clang depend target, no DESTDIR on HOSTCC build
ok patrick@

Pass CC and CXX to make depend as well. As soon as we support
dependencies in the clang makefiles we need to make sure to use
a compiler that supports C++11.

Adjust cross-toolchain environment to the changes in the clang build
fabric. Since we now also build and use LLVM's lld we can copy lld to
the proper directory akin to what we do with clang. Build compiler-rt
early as it is needed for building libc.

"go for it" mpi@

Export CROSSDIR so that the two users (ld.so and binutils) of that
variable actually work correctly. While there, adjust a bit for
consistency.

"nobody will complain if you break it ;)" kettenis@

We can now build parts of binutils on aarch64 so enable cross-binutils
again.

ok patrick@

Adjust cross-compilation infrastructure to cope with builduser
permission requirements.

From Dale Rahn.

ok tb@
"commit it" guenther@

With clang as cross-compiler we need to tell it where our cross
directory is so that the correct include and library paths are used.
For this we create a simple wrapper that calls clang in the correct
mode and also passes the sysroot.

ok kettenis@

Implement support for cross-compiling arm64 in Makefile.cross. This
adds code to build clang as cross-compiler instead of gcc. Since we
don't have binutils for aarch64 in base, skip the binutils cross-
toolchain build for that architecture.

ok kettenis@

Unhook sqlite3.

Retire zaurus, as it hasn't made the EABI jump and will be permanently broken
shortly when we use the hardware thread register in userland

armish handled some early-gen arm machines, which required tons of
workarounds. Some of them will soon stand in the way of armv7.
Off to the attic you go.

remove hppa64 port, which we never got going beyond broken single users.
hppa reverse-stack gives us a valuable test case, but most developers don't
have a 2nd one to proceed further with this.
ok kettenis

We are done providing support for the vax.
lots of agreement.

Restore parenthesis

Send solbourne to where the ...........

Send Aviion to same place as Nova II
discussed with jsg

libocurses can go to the Attic. last consumer of it was ramdisk more(1)
conceptual ok guenther millert nicm

Always honor ${DESTDIR}. OK miod@

Use target BSDOBJDIR when cross-compiling libraries

ok mpi@

Now that we do not need to build binutils 2.15 (except for gdb), simplify the
build logic.

Invert the binutils 2.17 logic to only list non-2.17 platforms.

Fix cross-build. OK jsing@ miod@ guenther@ deraadt@

Remove the last gzsig tendril

more NO_CROSS

lynx will go to ports. there are too many reasons for it to go there,
and not enough for it to stay. lengthy discussions.

Make sure to override PIE_DEFAULT when building a cross-gcc4

Remove more Kerberos cruft.

ok jca@ jmc@

Retire hp300, mvme68k and mvme88k ports. These ports have no users, keeping
this hardware alive is becoming increasingly difficult, and I should heed the
message sent by the three disks which have died on me over the last few days.

Noone sane will mourn these ports anyway. So long, and thanks for the fish.

Beagle has been replaced by armv7.

Remove support for COMPILER_VERSION == gcc2.
Change the logic depending upon COMPILER_VERSION everywhere, to assume gcc4
is the norm and to explicitely test for gcc3 when a different behaviour
is required.
No functional change intended. Be sure to `make install' in share/mk before
attempting to do anything.

tabs got lost in previous rev

ELF_TOOLCHAIN bye bye.

remove BIN{MODE,GRP,OWN} since bsd.own.mk is already included.

ok bmercer@

gumstix was never finished

the unused palm code goes away.

Allow the PIE/NOPIE variables in bsd.own.mk to be overriden. Force proper
values in Makefile.cross to make sure binutils are built with the proper
PIE_DEFAULT value for the target.
ok millert@ deraadt@

m68k ports are now ELF citizens.

Switch m88k ports to ELF.

Pass -DOPENBSD_CROSS to CFLAGS when cross-compiling gcc2. Temporary bandaid
until we get rid of it. Soon.

Make COMPILER_VERSION a definitive assignment in there, for it to override
<bsd.own.mk>. Allows cross-gcc to be built on a host using a different compiler
version than the target.

no need to know about afs, since it is gone

Garbage collect the USE_GCC variables; everything uses
COMPILER_VERSION now.

From Brad.

remove leftover NOLINT, WANTLINT, LINTFLAGS, LOBJ vars and lint targets.

ok guenther@

Add sqlite3 to NO_CROSS since it depends on libreadline - this would then
require cross-gnu to build before we can build cross-bin.

more mac68k bits for the attic

Do not try to cross build nginx, since it does not work.

- add avr32 which'll use gcc4/binutils-2.17
- remove BINUTILS217_ARCH from Makefile.cross as bsd.own.mk is included there
(requested by drahn@)

ok drahn@ deraadt@

Allow 64bit -> arm cross compilation as it appears safe (amd64 -> arm at least)
include bsd.own.mk to get the correct compiler architectures
(if only Makefile.cross is used)

Detect word-size mismatches between host and target, which result in
the cross compiler generating bad code, and abort

ok deraadt@

ia64 should use gcc4; sort the list of binutils-2.17 platforms while I'm there

ok jsing@, miod@

Switch hppa64 cross compiling environment to gcc4.

Switch hppa64 to binutils-2.17

remove mvmeppc; it is really rough shape. ok drahn miod

Allow switching between binutils and binutils 2.17.
NOTE: binutils-2.17 has not been tested on any archs. ok pirofti@

Typo, BINOWN -> BINGRP

ok miod@

unlink libsectok from the build; ok markus@ deraadt@

Copy the right bits to the right places for gcc4.

Switch mips-based ports to gcc4. Most of the work has been done by kettenis@.

Remove tn3270 reference.

add octeon as cross build target arch. ok miod@

nsd uses configure, so put it in NO_CROSS for now.

Sigh, check in the actual diff, not a backout of th previous change.
Use TARGET_ARCH, not MACHINE_ARCH. Fix from Tobias Ulmer. ok jsg@ jasper@

Use TARGET_ARCH, not MACHINE_ARCH. Fix from Tobias Ulmer. ok jsg@ jasper@

sync the list of gcc4 archs with bsd.own.mk
ok miod@

Fix cross build problem with cpp, marex pointed out that it wasn't committed.

Initial attempt at being COMPILER_VERSION and gcc4 compatible.

moko goes away.

Force MACHINE_CPU when building cross-toolchain or cross-building. This allows
loongson cross-tools to be configured correctly.

Add all arm based architectures. ok miod@.

Palm: Integration into the tree

OK deraadt@

Try to clean up the cross build environment, more of the tree cross builds now
ok miod@

socppc is powerpc

another cats remnant

feed bsdsrc/objdir and makeobjdir in cross-env allowing make obj and the like to work on random directories; found w/ rainerx0r; drahn@ ok

missing TARGET_ARCH handling (solbourne)

Early bits for a 32-bit SuperH toolchain, currently for little-endian
systems only; more work is necessary in ld land.

alpha is gcc3 now

typoo

System build pieces for armish arch.

Administrative changes for aviion, ok deraadt@

Install binutils objdump on a.out cross-tools; ok grange@

make sure cross-as and cross-ranlib get compiled again for a.out platforms.

photurisd is nomore; from Michael Knudsen

Disable lint for libraries.

ok deraadt@ espie@

Fix cross-etc-root-var, pick up target etc files, simplied fix from PR4511
ok grange@

Compute TARGET_ARCH in the Makefile, rather than in a shell, and use it
instead of MACHINE_ARCH to pick compiler and binutils versions.

looked at mickey@ grange@

Create proper symlink for a.out as.

Create a correct /usr/bin/cpp for a.out platforms.

telnetd is gone

Kill \n at EOF.

match gcc3 logic from bsd.own.mk and add a target_arch generation for arm

No need for cross nm/size in cross-env, our host tools are good enough.

ok mickey@

Unbreak cross-tools by providing TARGET_ARCH in include target
to get right config.status file for binutils and gcc.

ok mickey@

less confusing `TARGET_USE_GCC3' since it's only used locally.
Okay niklas@

- introduce TARGET_CANON/TARGET_ARCH shell variables, to make things slightly
clearer.

- don't put full path into symlinks that don't change directory.

- add a few NO_CROSS directories: stuff that depends on libcrypto, stuff
that needs host tools, and stuff that wants cross-configure.

okay niklas@

move cross-stuff into its own file.
okay mickey@, niklas@

Initial set of changes to build powerpc64 toolchain.

Disable PIC/PIE for powerpc64 for now.

Simplify by getting rid of the unecessary 'depend' rule.

ok patrick@

Add mips64el to LP64_ARCH.

OK patrick@

Re-sort the if-sequence which sets COMPILER_VERSION to be the same
as bsd.own.mk, otherwise on transitioning architectures the wrong
compiler is chosen. Also make sure BUILD_CLANG is forced for the
libcompiler_rt directory since cross-compiling with two possible
compilers would create too much headache.

Fix whitespace issues

powerpc cross-compilation on 64-bit hosts also appears to be safe.

Pass LINKER_VERSION to binutils and clang cross-build so that they
know who the system linker is.

Fix malformed conditional after socppc removal.

From Krystian Lewandowski

socppc makes an extended visit to the bigbucket.
ok kettenis

Don't assume that we use LLVM's ld(1) if we use clang(1) as main system
compiler. Only handle this case when LLD was actually installed during
the cross-compiler setup. Unbreaks armv7 after switching to clang since
it uses binutils' ld(1).

Compile the cross-compiler in the target's objdir instead of polluting
the host's. This also allows compiling multiple toolchains in parallel.
Clean up the clang object files afterwards so it doesn't collide with
the cross-compiled compiler build.

Fix cross-compiling GCC toolchains on Clang archs.

ok patrick@

Fix cross builds: no clang depend target, no DESTDIR on HOSTCC build
ok patrick@

Pass CC and CXX to make depend as well. As soon as we support
dependencies in the clang makefiles we need to make sure to use
a compiler that supports C++11.

Adjust cross-toolchain environment to the changes in the clang build
fabric. Since we now also build and use LLVM's lld we can copy lld to
the proper directory akin to what we do with clang. Build compiler-rt
early as it is needed for building libc.

"go for it" mpi@

Export CROSSDIR so that the two users (ld.so and binutils) of that
variable actually work correctly. While there, adjust a bit for
consistency.

"nobody will complain if you break it ;)" kettenis@

We can now build parts of binutils on aarch64 so enable cross-binutils
again.

ok patrick@

Adjust cross-compilation infrastructure to cope with builduser
permission requirements.

From Dale Rahn.

ok tb@
"commit it" guenther@

With clang as cross-compiler we need to tell it where our cross
directory is so that the correct include and library paths are used.
For this we create a simple wrapper that calls clang in the correct
mode and also passes the sysroot.

ok kettenis@

Implement support for cross-compiling arm64 in Makefile.cross. This
adds code to build clang as cross-compiler instead of gcc. Since we
don't have binutils for aarch64 in base, skip the binutils cross-
toolchain build for that architecture.

ok kettenis@

Unhook sqlite3.

Retire zaurus, as it hasn't made the EABI jump and will be permanently broken
shortly when we use the hardware thread register in userland

armish handled some early-gen arm machines, which required tons of
workarounds. Some of them will soon stand in the way of armv7.
Off to the attic you go.

remove hppa64 port, which we never got going beyond broken single users.
hppa reverse-stack gives us a valuable test case, but most developers don't
have a 2nd one to proceed further with this.
ok kettenis

We are done providing support for the vax.
lots of agreement.

Restore parenthesis

Send solbourne to where the ...........

Send Aviion to same place as Nova II
discussed with jsg

libocurses can go to the Attic. last consumer of it was ramdisk more(1)
conceptual ok guenther millert nicm

Always honor ${DESTDIR}. OK miod@

Use target BSDOBJDIR when cross-compiling libraries

ok mpi@

Now that we do not need to build binutils 2.15 (except for gdb), simplify the
build logic.

Invert the binutils 2.17 logic to only list non-2.17 platforms.

Fix cross-build. OK jsing@ miod@ guenther@ deraadt@

Remove the last gzsig tendril

more NO_CROSS

lynx will go to ports. there are too many reasons for it to go there,
and not enough for it to stay. lengthy discussions.

Make sure to override PIE_DEFAULT when building a cross-gcc4

Remove more Kerberos cruft.

ok jca@ jmc@

Retire hp300, mvme68k and mvme88k ports. These ports have no users, keeping
this hardware alive is becoming increasingly difficult, and I should heed the
message sent by the three disks which have died on me over the last few days.

Noone sane will mourn these ports anyway. So long, and thanks for the fish.

Beagle has been replaced by armv7.

Remove support for COMPILER_VERSION == gcc2.
Change the logic depending upon COMPILER_VERSION everywhere, to assume gcc4
is the norm and to explicitely test for gcc3 when a different behaviour
is required.
No functional change intended. Be sure to `make install' in share/mk before
attempting to do anything.

tabs got lost in previous rev

ELF_TOOLCHAIN bye bye.

remove BIN{MODE,GRP,OWN} since bsd.own.mk is already included.

ok bmercer@

gumstix was never finished

the unused palm code goes away.

Allow the PIE/NOPIE variables in bsd.own.mk to be overriden. Force proper
values in Makefile.cross to make sure binutils are built with the proper
PIE_DEFAULT value for the target.
ok millert@ deraadt@

m68k ports are now ELF citizens.

Switch m88k ports to ELF.

Pass -DOPENBSD_CROSS to CFLAGS when cross-compiling gcc2. Temporary bandaid
until we get rid of it. Soon.

Make COMPILER_VERSION a definitive assignment in there, for it to override
<bsd.own.mk>. Allows cross-gcc to be built on a host using a different compiler
version than the target.

no need to know about afs, since it is gone

Garbage collect the USE_GCC variables; everything uses
COMPILER_VERSION now.

From Brad.

remove leftover NOLINT, WANTLINT, LINTFLAGS, LOBJ vars and lint targets.

ok guenther@

Add sqlite3 to NO_CROSS since it depends on libreadline - this would then
require cross-gnu to build before we can build cross-bin.

more mac68k bits for the attic

Do not try to cross build nginx, since it does not work.

- add avr32 which'll use gcc4/binutils-2.17
- remove BINUTILS217_ARCH from Makefile.cross as bsd.own.mk is included there
(requested by drahn@)

ok drahn@ deraadt@

Allow 64bit -> arm cross compilation as it appears safe (amd64 -> arm at least)
include bsd.own.mk to get the correct compiler architectures
(if only Makefile.cross is used)

Detect word-size mismatches between host and target, which result in
the cross compiler generating bad code, and abort

ok deraadt@

ia64 should use gcc4; sort the list of binutils-2.17 platforms while I'm there

ok jsing@, miod@

Switch hppa64 cross compiling environment to gcc4.

Switch hppa64 to binutils-2.17

remove mvmeppc; it is really rough shape. ok drahn miod

Allow switching between binutils and binutils 2.17.
NOTE: binutils-2.17 has not been tested on any archs. ok pirofti@

Typo, BINOWN -> BINGRP

ok miod@

unlink libsectok from the build; ok markus@ deraadt@

Copy the right bits to the right places for gcc4.

Switch mips-based ports to gcc4. Most of the work has been done by kettenis@.

Remove tn3270 reference.

add octeon as cross build target arch. ok miod@

nsd uses configure, so put it in NO_CROSS for now.

Sigh, check in the actual diff, not a backout of th previous change.
Use TARGET_ARCH, not MACHINE_ARCH. Fix from Tobias Ulmer. ok jsg@ jasper@

Use TARGET_ARCH, not MACHINE_ARCH. Fix from Tobias Ulmer. ok jsg@ jasper@

sync the list of gcc4 archs with bsd.own.mk
ok miod@

Fix cross build problem with cpp, marex pointed out that it wasn't committed.

Initial attempt at being COMPILER_VERSION and gcc4 compatible.

moko goes away.

Force MACHINE_CPU when building cross-toolchain or cross-building. This allows
loongson cross-tools to be configured correctly.

Add all arm based architectures. ok miod@.

Palm: Integration into the tree

OK deraadt@

Try to clean up the cross build environment, more of the tree cross builds now
ok miod@

socppc is powerpc

another cats remnant

feed bsdsrc/objdir and makeobjdir in cross-env allowing make obj and the like to work on random directories; found w/ rainerx0r; drahn@ ok

missing TARGET_ARCH handling (solbourne)

Early bits for a 32-bit SuperH toolchain, currently for little-endian
systems only; more work is necessary in ld land.

alpha is gcc3 now

typoo

System build pieces for armish arch.

Administrative changes for aviion, ok deraadt@

Install binutils objdump on a.out cross-tools; ok grange@

make sure cross-as and cross-ranlib get compiled again for a.out platforms.

photurisd is nomore; from Michael Knudsen

Disable lint for libraries.

ok deraadt@ espie@

Fix cross-etc-root-var, pick up target etc files, simplied fix from PR4511
ok grange@

Compute TARGET_ARCH in the Makefile, rather than in a shell, and use it
instead of MACHINE_ARCH to pick compiler and binutils versions.

looked at mickey@ grange@

Create proper symlink for a.out as.

Create a correct /usr/bin/cpp for a.out platforms.

telnetd is gone

Kill \n at EOF.

match gcc3 logic from bsd.own.mk and add a target_arch generation for arm

No need for cross nm/size in cross-env, our host tools are good enough.

ok mickey@

Unbreak cross-tools by providing TARGET_ARCH in include target
to get right config.status file for binutils and gcc.

ok mickey@

less confusing `TARGET_USE_GCC3' since it's only used locally.
Okay niklas@

- introduce TARGET_CANON/TARGET_ARCH shell variables, to make things slightly
clearer.

- don't put full path into symlinks that don't change directory.

- add a few NO_CROSS directories: stuff that depends on libcrypto, stuff
that needs host tools, and stuff that wants cross-configure.

okay niklas@

move cross-stuff into its own file.
okay mickey@, niklas@

Simplify by getting rid of the unecessary 'depend' rule.

ok patrick@

Add mips64el to LP64_ARCH.

OK patrick@

Re-sort the if-sequence which sets COMPILER_VERSION to be the same
as bsd.own.mk, otherwise on transitioning architectures the wrong
compiler is chosen. Also make sure BUILD_CLANG is forced for the
libcompiler_rt directory since cross-compiling with two possible
compilers would create too much headache.

Fix whitespace issues

powerpc cross-compilation on 64-bit hosts also appears to be safe.

Pass LINKER_VERSION to binutils and clang cross-build so that they
know who the system linker is.

Fix malformed conditional after socppc removal.

From Krystian Lewandowski

socppc makes an extended visit to the bigbucket.
ok kettenis

Don't assume that we use LLVM's ld(1) if we use clang(1) as main system
compiler. Only handle this case when LLD was actually installed during
the cross-compiler setup. Unbreaks armv7 after switching to clang since
it uses binutils' ld(1).

Compile the cross-compiler in the target's objdir instead of polluting
the host's. This also allows compiling multiple toolchains in parallel.
Clean up the clang object files afterwards so it doesn't collide with
the cross-compiled compiler build.

Fix cross-compiling GCC toolchains on Clang archs.

ok patrick@

Fix cross builds: no clang depend target, no DESTDIR on HOSTCC build
ok patrick@

Pass CC and CXX to make depend as well. As soon as we support
dependencies in the clang makefiles we need to make sure to use
a compiler that supports C++11.

Adjust cross-toolchain environment to the changes in the clang build
fabric. Since we now also build and use LLVM's lld we can copy lld to
the proper directory akin to what we do with clang. Build compiler-rt
early as it is needed for building libc.

"go for it" mpi@

Export CROSSDIR so that the two users (ld.so and binutils) of that
variable actually work correctly. While there, adjust a bit for
consistency.

"nobody will complain if you break it ;)" kettenis@

We can now build parts of binutils on aarch64 so enable cross-binutils
again.

ok patrick@

Adjust cross-compilation infrastructure to cope with builduser
permission requirements.

From Dale Rahn.

ok tb@
"commit it" guenther@

With clang as cross-compiler we need to tell it where our cross
directory is so that the correct include and library paths are used.
For this we create a simple wrapper that calls clang in the correct
mode and also passes the sysroot.

ok kettenis@

Implement support for cross-compiling arm64 in Makefile.cross. This
adds code to build clang as cross-compiler instead of gcc. Since we
don't have binutils for aarch64 in base, skip the binutils cross-
toolchain build for that architecture.

ok kettenis@

Unhook sqlite3.

Retire zaurus, as it hasn't made the EABI jump and will be permanently broken
shortly when we use the hardware thread register in userland

armish handled some early-gen arm machines, which required tons of
workarounds. Some of them will soon stand in the way of armv7.
Off to the attic you go.