Remove obsolete comment.

branches: 1.1.1;
Initial revision

Disable DEBUG options properly.

Remove trailing spaces and TABs.

branches: 1.29.18; 1.29.44;
centralize vers.c building for standalone programs.

branches: 1.28.6; 1.28.10; 1.28.14;
Add empty LIBCRTI= as LIBCRT0 to build sa programs without installed DESTDIR.

XXX: probabry we should have bsd.saprog.mk or something.

Use <bsd.klinks.mk>
Add ${_MKTARGET_LINK}

branches: 1.26.2;
Revert previous. This boot loader has its own version number,
independent of the kernel version.

Add references to ${_NETBSD_VERSION_DEPENDS} for files that
need to be re-created when the NetBSD version changes. They
will also be re-created when any build settings are changed.

branches: 1.24.4; 1.24.14;
Drop bootprog_maker (formerly enabled by -M) and bootprog_date (formerly
disabled by -D) from the output of newvers_stand.sh. Change -D to the
inverted logic, so that it adds the date to bootprog_rev in ().

Change all platforms accordingly. -D is added if MKREPRO is not yes and
wasn't present before. Platforms that didn't use -D don't depend on
MKREPRO now either.

branches: 1.23.4; 1.23.6; 1.23.8;
Don't abuse INSECURE for allowing exec() to load files not owned by uid 0.

Adjust references where this has been used.

- WARNSfy
- make sure machine and ${MACHINE_ARCH} symlinks are always created
- add incpath to AFLAGS too for build without DESTDIR

branches: 1.21.2;
Use ${TOOL_AWK} instead of ${AWK} or plain "awk" in make commands.
Pass AWK=${TOOL_AWK:Q} to shell scripts that use awk.

branches: 1.20.74; 1.20.78; 1.20.84;
merge ktrace-lwp.

Use ${HOST_SH} instead of `sh'.

If necessary, pull in <bsd.sys.mk> to get the definition of HOST_SH;
Makefiles that pull in one of (most of) <bsd.*.mk> will get this anyway.

Define LIB{CRT0,C,CRTBEGIN,CRTEND} as nothing, we don't need to link
against them.

NOxxx= variables should appear first

branches: 1.16.6;
Determine turbo-ness based on the ROM machine type here, too.

Comprehensive patches from Christian Limpach:
* Fix problems with the DMA and SCSI drivers.
* Make turbo machines sort of work.
Additional fixes from me:
* Determine if we're a turbo at boot time, by looking at the ROM machine type.
* Set the display size correctly (1120 pixels wide, but padded to 1152 only on
non-turbo machines).
Caveats:
* SCSI doesn't work on the turbo (or at least it blows chunks with no devices
attached).
* Media selection doesn't work on the turbo (the BMAP stuff doesn't exist on
turbo machines).
* The boot block is prone to timing out.

Apply patches from Christian Limpach:

- NeXT label reading support
- SCSI dma fixes
- media support for if_xe.c

Some of these need more cleanup, but at least make SCSI support usable on
the NeXT.

branches: 1.13.2;
don't strip the resulting bootprog on install; it's raw binary
XXX I hope this is correct way to do this

use ${.OBJDIR} rather than ${.CURDIR}/obj.${MACHINE} (eek)
g/c -I${.CURDIR} and -I${S}/lib/libkern from ${INCLUDES}
build libsa dev_net.c via .PATH and SRCS rather than special rules

add -ffreestanding to CFLAGS, so that this wouldn't attempt to link in libc(ugh)
g/c the -Wno-main and HAVE_GCC28 stuff
make the machine symlinks via common .BEGIN cookie, so that they would
be properly setup for 'depend' or 'dependall' targets too; g/c the limits.h
symlink, which doesn't seem to be needed

MKfoo=no -> NOfoo

branches: 1.9.2; 1.9.8;
switch next68k to ELF. highlight include:
add "%" prefix to register names in assembly code.
change assembly functions to return pointer values in %a0 instead of %d0.
C symbols no longer prepend an underscore, adjust assembly code for this.
32-bit values are now 32-bit aligned instead of 16-bit aligned,
adjust structure packing and padding to override this where necessary.
make EXEC_ELF std, make EXEC_AOUT and COMPAT_AOUT_M68K optional.
use the MI loadfile() instead of several home-grown versions.

branches: 1.8.2;
do bootblock versioning standard way via sys/conf/newvers_stand.sh, keep what
was previously called "subversion" as "build", remove old cruft from newvers.sh

XXX very lighly tested build on NetBSD/hp300 1.4.1, though could not do full
XXX build due to toolchain differences to -current

branches: 1.7.8;
tweaks to compile with egcs -Werror
minor changes to scsi driver register accesses.

convert from NOxxx= to MKxxx=no.
include <bsd.own.mk> if testing a MKxxx variable.

Use WARNS=1 to get -Wall, and properly disable -Wmain on gcc 2.8

added -Wno-main if USE_EGCS is defined

Commented out some bus_dma code, until I can fix.
(bus_dma.c needs sync with alpha port!)
Continued progress on scsi driver.
A couple of other compiler warning level of tweaks.

Removed unnecessary bug-workaround headers.
Minor compilation and header tweak.

branches: 1.1.1;
Initial revision

Remove trailing spaces and TABs.

branches: 1.1.1;
Initial revision

Replace DELAY() with one in hp300 bootloader and adjust cpuspeed counts.

Fix boot failure on my ancient Seagate ST52160N drive.
It looks some of such old drives can't respond to SCSI
test-unit-ready command without proper wait after SCSI bus reset.
Bump version again to denote a fix.

XXX we should re-evaluate cpuspeed counts for DELAY() in bootloaders
(where cache is disabled) on other m68k ports, hp300 and luna68k etc.

Use common declarations and macros in proper headers.

Also fix inconsistent sdopen() and sdstrategy() args and
remove useless #if 0'ed out code.
No functional change.

Add and check machine type NeXT_CUBE_TURBO (type 8).

Info from Andreas Grabher on port-next68k@.

branches: 1.12.22; 1.12.48;
PR 51200 gets in libsa considered harmful: use kgets

branches: 1.11.6;
kill sprintf

branches: 1.10.112; 1.10.122; 1.10.128;
merge ktrace-lwp.

de-__P, remove register, ansify.

branches: 1.8.8;
Fix NULL change lossage.

branches: 1.7.6;
Determine turbo-ness based on the ROM machine type here, too.

Comprehensive patches from Christian Limpach:
* Fix problems with the DMA and SCSI drivers.
* Make turbo machines sort of work.
Additional fixes from me:
* Determine if we're a turbo at boot time, by looking at the ROM machine type.
* Set the display size correctly (1120 pixels wide, but padded to 1152 only on
non-turbo machines).
Caveats:
* SCSI doesn't work on the turbo (or at least it blows chunks with no devices
attached).
* Media selection doesn't work on the turbo (the BMAP stuff doesn't exist on
turbo machines).
* The boot block is prone to timing out.

Apply patches from Christian Limpach:

- NeXT label reading support
- SCSI dma fixes
- media support for if_xe.c

Some of these need more cleanup, but at least make SCSI support usable on
the NeXT.

branches: 1.4.2; 1.4.8; 1.4.16;
switch next68k to ELF. highlight include:
add "%" prefix to register names in assembly code.
change assembly functions to return pointer values in %a0 instead of %d0.
C symbols no longer prepend an underscore, adjust assembly code for this.
32-bit values are now 32-bit aligned instead of 16-bit aligned,
adjust structure packing and padding to override this where necessary.
make EXEC_ELF std, make EXEC_AOUT and COMPAT_AOUT_M68K optional.
use the MI loadfile() instead of several home-grown versions.

branches: 1.3.2;
do bootblock versioning standard way via sys/conf/newvers_stand.sh, keep what
was previously called "subversion" as "build", remove old cruft from newvers.sh

XXX very lighly tested build on NetBSD/hp300 1.4.1, though could not do full
XXX build due to toolchain differences to -current

branches: 1.2.14;
Removed unnecessary bug-workaround headers.
Minor compilation and header tweak.

branches: 1.1.1;
Initial revision

Use common declarations and macros in proper headers.

Also fix inconsistent sdopen() and sdstrategy() args and
remove useless #if 0'ed out code.
No functional change.

branches: 1.7.170; 1.7.196;
merge ktrace-lwp.

Use FS_OPS() macro.

de-__P, remove register, ansify.

branches: 1.4.8;
Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22364, verified by myself.

branches: 1.3.2;
cope with removal of this sick NENTS macro from libsa/netif.h

Apply patches from Christian Limpach:

- NeXT label reading support
- SCSI dma fixes
- media support for if_xe.c

Some of these need more cleanup, but at least make SCSI support usable on
the NeXT.

branches: 1.1.1;
Initial revision

tweaks to compile with egcs -Werror
minor changes to scsi driver register accesses.

branches: 1.1.1;
Initial revision

tweaks to compile with egcs -Werror
minor changes to scsi driver register accesses.

branches: 1.1.1;
Initial revision

Use common declarations and macros in proper headers.

Also fix inconsistent sdopen() and sdstrategy() args and
remove useless #if 0'ed out code.
No functional change.

Remove trailing spaces and TABs.

branches: 1.6.32; 1.6.58;
Unify all arch/*/stand's atoi() to MI libsa.
lib/libsa/atoi.c was separated from lib/libsa/bootcfg.c.
PR/49084

branches: 1.5.122;
merge ktrace-lwp.

PR/21172: Christian Biere: ctype function misuse.

de-__P, remove register, ansify.

branches: 1.2.42; 1.2.50;
tweaks to compile with egcs -Werror
minor changes to scsi driver register accesses.

branches: 1.1.1;
Initial revision

Make sure to specify volatile explicitly on DMA register accesses.

It looks booting from SCSI disks on next68k have been broken
since NetBSD 1.6 days, but now it works.
Should be pulled up to netbsd-10 and netbsd-9.

branches: 1.3.202; 1.3.228;
Comprehensive patches from Christian Limpach:
* Fix problems with the DMA and SCSI drivers.
* Make turbo machines sort of work.
Additional fixes from me:
* Determine if we're a turbo at boot time, by looking at the ROM machine type.
* Set the display size correctly (1120 pixels wide, but padded to 1152 only on
non-turbo machines).
Caveats:
* SCSI doesn't work on the turbo (or at least it blows chunks with no devices
attached).
* Media selection doesn't work on the turbo (the BMAP stuff doesn't exist on
turbo machines).
* The boot block is prone to timing out.

Apply patches from Christian Limpach:

- NeXT label reading support
- SCSI dma fixes
- media support for if_xe.c

Some of these need more cleanup, but at least make SCSI support usable on
the NeXT.

branches: 1.1.1;
Initial revision

Use common declarations and macros in proper headers.

Also fix inconsistent sdopen() and sdstrategy() args and
remove useless #if 0'ed out code.
No functional change.

branches: 1.19.6; 1.19.32;
fix various gcc6 m68k issues:
- bad indentation. next68k en.c and mvme68k le_poll.c fixes real issues
in error handling, the rest are NFCI.
- pass 68030 flags as appropriate for mvme68k.
- next68k nextrom.c has -Warray-bounds ignored for an odd expression
that appears to run before relocation, and needs manual offsets
added which trips bounds array checking.

with this all m68k ports build with GCC 6.

branches: 1.18.16;
fix build with gcc48.

branches: 1.17.22; 1.17.38; 1.17.40;
kill extra whitespaces
reviewed by tsutsui@

bcopy -> memcpy

branches: 1.15.2;
Replace time_t values in libsa sources with the following two types
to avoid unnecessary 64 bit ops which would make binaries larger:

satime_t (currently unsigned int):
numbers in seconds returned by the machine dependent getsecs() function
which are used to measure relative time

saseconds_t (currently int):
numbers in seconds used to specify timeout to network drivers

Per discussion on current-users.

branches: 1.14.74; 1.14.78; 1.14.86;
merge ktrace-lwp.

Cosmetic changes, notably drop trailing spaces.

Yes, it was a cool trick >20 years ago to use "0123456789abcdef"[a] to
implement, xtoa(), but I think defining the samestring 50 times is a bit
too much. Defined HEXDIGITS and hexdigits in subr_prf.c and use it...

de-__P, remove register, ansify.

branches: 1.10.2; 1.10.10;
DMA, not dma nor Dma.

cope with removal of this sick NENTS macro from libsa/netif.h

Don't blindly receive all multicast packets just because we're on a turbo.
Pay lip service to making promiscuous mode work.

In en_put(), if we don't get a packet, just return 0. This is how we tell
the libsa code to retransmit.

Determine turbo-ness based on the ROM machine type here, too.

Comprehensive patches from Christian Limpach:
* Fix problems with the DMA and SCSI drivers.
* Make turbo machines sort of work.
Additional fixes from me:
* Determine if we're a turbo at boot time, by looking at the ROM machine type.
* Set the display size correctly (1120 pixels wide, but padded to 1152 only on
non-turbo machines).
Caveats:
* SCSI doesn't work on the turbo (or at least it blows chunks with no devices
attached).
* Media selection doesn't work on the turbo (the BMAP stuff doesn't exist on
turbo machines).
* The boot block is prone to timing out.

Apply patches from Christian Limpach:

- NeXT label reading support
- SCSI dma fixes
- media support for if_xe.c

Some of these need more cleanup, but at least make SCSI support usable on
the NeXT.

branches: 1.3.16; 1.3.20; 1.3.28;
include <lib/libkern/libkern.h> for intoa()/inet_ntoa()

branches: 1.2.4;
tweaks to compile with egcs -Werror
minor changes to scsi driver register accesses.

branches: 1.1.1;
Initial revision

Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.3.22; 1.3.60;
Determine turbo-ness based on the ROM machine type here, too.

Comprehensive patches from Christian Limpach:
* Fix problems with the DMA and SCSI drivers.
* Make turbo machines sort of work.
Additional fixes from me:
* Determine if we're a turbo at boot time, by looking at the ROM machine type.
* Set the display size correctly (1120 pixels wide, but padded to 1152 only on
non-turbo machines).
Caveats:
* SCSI doesn't work on the turbo (or at least it blows chunks with no devices
attached).
* Media selection doesn't work on the turbo (the BMAP stuff doesn't exist on
turbo machines).
* The boot block is prone to timing out.

branches: 1.1.1;
Initial revision

Apply patches from Christian Limpach:

- NeXT label reading support
- SCSI dma fixes
- media support for if_xe.c

Some of these need more cleanup, but at least make SCSI support usable on
the NeXT.

branches: 1.1.1;
Initial revision

Use common declarations and macros in proper headers.

Also fix inconsistent sdopen() and sdstrategy() args and
remove useless #if 0'ed out code.
No functional change.

Remove trailing spaces and TABs.

branches: 1.7.102; 1.7.128;
Remove __attribute__((__noreturn__)) from things already marked __dead
Found by the department of redundancy department.

branches: 1.6.50; 1.6.60; 1.6.64;
merge ktrace-lwp.

branches: 1.5.8;
de-__P, remove register, ansify.

branches: 1.4.6; 1.4.14;
Apply patches from Christian Limpach:

- NeXT label reading support
- SCSI dma fixes
- media support for if_xe.c

Some of these need more cleanup, but at least make SCSI support usable on
the NeXT.

branches: 1.3.2; 1.3.8; 1.3.16;
switch next68k to ELF. highlight include:
add "%" prefix to register names in assembly code.
change assembly functions to return pointer values in %a0 instead of %d0.
C symbols no longer prepend an underscore, adjust assembly code for this.
32-bit values are now 32-bit aligned instead of 16-bit aligned,
adjust structure packing and padding to override this where necessary.
make EXEC_ELF std, make EXEC_AOUT and COMPAT_AOUT_M68K optional.
use the MI loadfile() instead of several home-grown versions.

branches: 1.2.20;
tweaks to compile with egcs -Werror
minor changes to scsi driver register accesses.

branches: 1.1.1;
Initial revision

Removed unnecessary bug-workaround headers.
Minor compilation and header tweak.

branches: 1.1.1;
Initial revision

Convert TNF licenses to new 2 clause variant

branches: 1.2.126; 1.2.128; 1.2.130;
do bootblock versioning standard way via sys/conf/newvers_stand.sh, keep what
was previously called "subversion" as "build", remove old cruft from newvers.sh

XXX very lighly tested build on NetBSD/hp300 1.4.1, though could not do full
XXX build due to toolchain differences to -current

branches: 1.1.1;
Initial revision

Use common declarations and macros in proper headers.

Also fix inconsistent sdopen() and sdstrategy() args and
remove useless #if 0'ed out code.
No functional change.

Remove trailing spaces and TABs.

branches: 1.6.70; 1.6.96;
Replace time_t values in libsa sources with the following two types
to avoid unnecessary 64 bit ops which would make binaries larger:

satime_t (currently unsigned int):
numbers in seconds returned by the machine dependent getsecs() function
which are used to measure relative time

saseconds_t (currently int):
numbers in seconds used to specify timeout to network drivers

Per discussion on current-users.

branches: 1.5.64; 1.5.68; 1.5.76;
Fix clock reading problem on turbo nexts and color slabs. First
reported by: Timm Wetzel twetzel at gwdg.de on 2001-06-22, and now
again by: Cory Bajus cbajus at mts.net

branches: 1.4.4; 1.4.6; 1.4.8; 1.4.10;
merge ktrace-lwp.

branches: 1.3.8;
de-__P, remove register, ansify.

branches: 1.2.42; 1.2.50;
tweaks to compile with egcs -Werror
minor changes to scsi driver register accesses.

branches: 1.1.1;
Initial revision

branches: 1.2.2; 1.2.4;
Replace DELAY() with one in hp300 bootloader and adjust cpuspeed counts.

Fix boot failure on my ancient Seagate ST52160N drive.
It looks some of such old drives can't respond to SCSI
test-unit-ready command without proper wait after SCSI bus reset.
Bump version again to denote a fix.

XXX we should re-evaluate cpuspeed counts for DELAY() in bootloaders
(where cache is disabled) on other m68k ports, hp300 and luna68k etc.

Use common declarations and macros in proper headers.

Also fix inconsistent sdopen() and sdstrategy() args and
remove useless #if 0'ed out code.
No functional change.

s/bufffer/buffer/ in comments and log message.

Replace DELAY() with one in hp300 bootloader and adjust cpuspeed counts.

Fix boot failure on my ancient Seagate ST52160N drive.
It looks some of such old drives can't respond to SCSI
test-unit-ready command without proper wait after SCSI bus reset.
Bump version again to denote a fix.

XXX we should re-evaluate cpuspeed counts for DELAY() in bootloaders
(where cache is disabled) on other m68k ports, hp300 and luna68k etc.

Use common declarations and macros in proper headers.

Also fix inconsistent sdopen() and sdstrategy() args and
remove useless #if 0'ed out code.
No functional change.

Disable DEBUG options properly.

Remove trailing spaces and TABs.

branches: 1.10.96; 1.10.122;
Convert from using bcopy() to memcpy(), so that this builds again.

branches: 1.9.40;
Follow Izumi Tsutsui's advice, and use uint8_t* rather than char*.

Need char* for pointer arithmetic and array indexing.

Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.6.18; 1.6.56;
DMA, not dma nor Dma.

Ignore bus exception errors here.
XXX I'm not sure exactly why this is necessary...

Apply patches from Christian Limpach:

- NeXT label reading support
- SCSI dma fixes
- media support for if_xe.c

Some of these need more cleanup, but at least make SCSI support usable on
the NeXT.

branches: 1.3.2;
some esp register names are no longer provided by next68k espreg.h,
use MI <dev/ic/ncr53c9xreg.h> ones

branches: 1.2.22; 1.2.26;
tweaks to compile with egcs -Werror
minor changes to scsi driver register accesses.

branches: 1.1.1;
Initial revision

src is too big these days to tolerate superfluous apostrophes. It's
"its", people!

branches: 1.4.88;
Follow Izumi Tsutsui's advice, and use uint8_t* rather than char*.

Need char* for pointer arithmetic and array indexing.

Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.1.1;
Initial revision

Use common declarations and macros in proper headers.

Also fix inconsistent sdopen() and sdstrategy() args and
remove useless #if 0'ed out code.
No functional change.

branches: 1.1.1;
Initial revision

Use common declarations and macros in proper headers.

Also fix inconsistent sdopen() and sdstrategy() args and
remove useless #if 0'ed out code.
No functional change.

Disable DEBUG options properly.

Request only 36 bytes for a response of INQUIRY command for legacy drives.

Some drives don't respond larger requested size for newer
SCSI3 devices and not all drivers can handle short xfers.
We should fix drivers to handle such short xfers properly,
but we need only SCSI device type here (and the 36 bytes are
enough even if we want vendor and product names on a bootloader).

The problem is reported from Andreas Grabher (a maintainer of NeXT
Computer Emulator)a on port-next68k@:
https://mail-index.netbsd.org/port-next68k/2023/02/thread1.html

Avoid possible division by zero trap in error cases to make debug easier.

Remove trailing spaces and TABs.

branches: 1.12.34; 1.12.60;
make this compile (missing header file)

branches: 1.11.12; 1.11.22; 1.11.26;
bzero -> memset

branches: 1.10.62; 1.10.70; 1.10.76;
gcc4 fix: correct of couple of casting errors picked up by gcc4.

branches: 1.9.2; 1.9.6;
free -> dealloc
unsigned -> size_t for alloc/dealloc

branches: 1.8.2;
merge ktrace-lwp.

branches: 1.7.4;
Part 1 of a cleanup pass over the SCSI subsystem. The aim is to name
everything "scsi_*", since we really are talking about the SCSI command
set, ATAPI transport not withstanding. Improve the names of many structures,
and prepend "SCSI_" onto all SCSI command opcodes. Place items described
by the SCSI Primary Commands document into scsi_spc.h.

branches: 1.6.2; 1.6.4;
READ_CAPACITY -> READ_CAPACITY_10

USe more appropriate macro/struct names for READ/WRITE (6) and
READ/WRITE (10).

move structs nextstep_disklabel/cpu_partition and appendant #defines
to sys/sys/bootblock.h
- rename to next68k_disklabel and next68k_partition
- use {u,}int{8,16,32}_t instead of char/short/int (suggested by D. Laight)

branches: 1.3.6;
Apply patches from Christian Limpach:

- NeXT label reading support
- SCSI dma fixes
- media support for if_xe.c

Some of these need more cleanup, but at least make SCSI support usable on
the NeXT.

branches: 1.2.22; 1.2.26; 1.2.34;
tweaks to compile with egcs -Werror
minor changes to scsi driver register accesses.

branches: 1.1.1;
Initial revision

Remove trailing spaces and TABs.

branches: 1.2.8; 1.2.218; 1.2.244;
switch next68k to ELF. highlight include:
add "%" prefix to register names in assembly code.
change assembly functions to return pointer values in %a0 instead of %d0.
C symbols no longer prepend an underscore, adjust assembly code for this.
32-bit values are now 32-bit aligned instead of 16-bit aligned,
adjust structure packing and padding to override this where necessary.
make EXEC_ELF std, make EXEC_AOUT and COMPAT_AOUT_M68K optional.
use the MI loadfile() instead of several home-grown versions.

branches: 1.1.1;
Initial revision

Removed unnecessary bug-workaround headers.
Minor compilation and header tweak.

branches: 1.1.1;
Initial revision

Actually bump version (missed in the previous commit).

Bump version again to denote NeXT_CUBE_TURBO support.

Bump version to 1.6 to denote recent bootloader's >20 years old bug fixes.

All these fixes should be pulled up to netbsd-10 and netbsd-9.

branches: 1.4.2; 1.4.212; 1.4.238;
Bump the version numbers of all bootloaders which use loadfile_elfXX() now
that it tries hard to avoid backwards seeks.

branches: 1.3.4;
Bump version numbers of all bootloaders that use loadfile/ELF, to account for
my changes to symbol loading. I should probably have done this at the time,
but it's better late than never.

branches: 1.2.2;
switch next68k to ELF. highlight include:
add "%" prefix to register names in assembly code.
change assembly functions to return pointer values in %a0 instead of %d0.
C symbols no longer prepend an underscore, adjust assembly code for this.
32-bit values are now 32-bit aligned instead of 16-bit aligned,
adjust structure packing and padding to override this where necessary.
make EXEC_ELF std, make EXEC_AOUT and COMPAT_AOUT_M68K optional.
use the MI loadfile() instead of several home-grown versions.

branches: 1.1.2; 1.1.4;
do bootblock versioning standard way via sys/conf/newvers_stand.sh, keep what
was previously called "subversion" as "build", remove old cruft from newvers.sh

XXX very lighly tested build on NetBSD/hp300 1.4.1, though could not do full
XXX build due to toolchain differences to -current