This only affects the -current early adopters and developers who have
done a 'make world' in the last few weeks and as a result installed a
gcc-3.1 version of /usr/bin/c++ but without the corresponding library
support that this now requires. This is a temporary hack that should be
deleted within a few weeks. In this case we will use the existing
gperf/groff one last time around for the early stage1 bootstrap. (This
isn't so bad, because we were unconditionally using the host one before)
Rename `incsinstall' to `installincludes'.
Make `includes' a -j safe shortcut for `buildincludes' + `installincludes'.
`buildincludes' and `installincludes' are SUBDIR friendly, if run directly.
via INCS. Implemented INCSLINKS (equivalent to SYMLINKS) to
handle symlinking include files. Allow for multiple groups of
include files to be installed, with the powerful INCSGROUPS knob.
Documentation to follow.
Added standard `includes' and `incsinstall' targets, use them
in Makefile.inc1. Headers from the following makefiles were
not installed before (during `includes' in Makefile.inc1):
kerberos5/lib/libtelnet/Makefile
lib/libbz2/Makefile
lib/libdevinfo/Makefile
lib/libform/Makefile
lib/libisc/Makefile
lib/libmenu/Makefile
lib/libmilter/Makefile
lib/libpanel/Makefile
Replaced all `beforeinstall' targets for installing includes
with the INCS stuff.
Renamed INCDIR to INCSDIR, for consistency with FILES and SCRIPTS,
and for compatibility with NetBSD. Similarly for INCOWN, INCGRP,
and INCMODE.
Consistently use INCLUDEDIR instead of /usr/include.
gnu/lib/libstdc++/Makefile and gnu/lib/libsupc++/Makefile changes
were only lightly tested due to the missing contrib/libstdc++-v3.
I fully tested the pre-WIP_GCC31 version of this patch with the
contrib/libstdc++.295 stuff.
These changes have been tested on i386 with the -DNO_WERROR "make
world" and "make release".
You need to set TARGET_ARCH and possibly TARGET, the same
way you normally do it for a cross build(7).
Renamed `distribworld' to a more natutal `distributeworld'.
Put pwd_mkdb(8) under ${INSTALLTMP}; for `distributeworld'.
Do not install games and profiled libraries to the ${CHROOTDIR}
with the initial installworld.
Eliminate the need in the second installworld. For that, make sure
_everything_ is built in the "world" environment, using the right
tool chain.
Added SUBDIR_OVERRIDE helper stuff to Makefile.inc1. Split the
buildworld process into stages, and skip some stages when
SUBDIR_OVERRIDE is set (used to build crypto, krb4, and krb5
dists).
Added NO_MAKEDB_RUN knob to Makefile.inc1 to avoid running
makewhatis(1) at the end of installworld (used when making crypto,
krb4, and krb5 dists).
In release/scripts/doFS.sh, ensure that the correct boot blocks are
used.
Moved the creation of the "crypto" dist from release.5 to
release.2.
In release.3 and doMFSKERN, build kernels in the "world"
environment. KERNELS now means "additional" kernels, GENERIC is
always built.
Ensure we build crunched binaries in the "world" environment.
Obfuscate release/Makefile some more (WMAKEENV) to achieve this.
Inline createBOOTMFS target.
Use already built GENERIC kernel modules to augment mfsfd's
/stand/modules. GC doMODULES as such.
Assorted fixes:
Get rid of the "afterdistribute" target by moving the single use
of it from sys/Makefile to etc/Makefile's "distribute".
Makefile.inc1: apparently "etc" no longer needs to be last for
"distribute" to succeed.
gnu/usr.bin/perl/library/Makefile.inc: do not override the
"install" and "distribute" targets, do it the "canonical" way.
release/scripts/{man,cat}pages-make.sh: make sure Perl manpages and
catpages appear in the right dists. Note that because Perl does
not respect the MANBUILDCAT (and NOMAN), this results in a loss of
/usr/share/perl/man/cat* empty directories. This will be fixed
soon.
Turn MAKE_KERBEROS4 into a plain boolean variable (if it is set it
means "make KerberosIV"), as documented in the make.conf(5)
manpage. Most of the userland makefiles did not test it for "YES"
anyway.
XXX Should specialized kerberized libpam versions be included into
the krb4 and krb5 dists? (libpam.a would be incorrect anyway if
both krb4 and krb5 dists were choosen.)
Make sure "games" dist is made before "catpages", otherwise games
catpages settle in the wrong dist.
Fast build machine provided by: Igor Kucherenko <kivvy@sunbay.com>
and UPDATING and has been posted to both freebsd-current and
freebsd-stable, users are still not adding the required smmsp user and
group before doing an installworld. Therefore, don't let users do an
installworld unless they have followed directions.
Add a new installcheck Makefile target which installworld runs before
actually starting the installation. This target can be used by other parts
of userland as well. The first addition to the target is to check for the
smmsp user and group if NO_SENDMAIL isn't defined.
Others may add checks to this target as they see fit.
MFC after: 1 week
breakage with ioctl.c. The .depend file should track dependencies
just fine, and the worst we can have is to miss new ioctls.
But I still think it's a good idea to have -DNOCLEAN build produce
the same ioctl.c as it would without -DNOCLEAN.
Prodded for a long time by: bde
of setting ${TARGET} -- make it default to ${MACHINE} if we are
not cross-building, and ${TARGET_ARCH} otherwise.
Set MAKEOBJDIREPREFIX based on ${TARGET}, not on ${TARGET_ARCH}.
This is useful if you want to cross-build pc98 worlds on i386.
Install sys/<arch>/include/pc/*.h to /usr/include/machine/pc/.
PR: docs/29534
Install sys/netatm/*/*.h to /usr/include/netatm/*/.
Don't install compatibility symlinks for <machine/soundcard.h>
and <machine/joystick.h>. Three years is enough to be aware of
the change, and these weren't visible in the SHARED=symlinks
case.
Back out include/Makefile,v 1.160 that was a null change anyway
due to the bug in the path, and we now don't want to install
these headers because they would otherwise be invisible in the
SHARED=symlinks case.
Don't install IPFILTER headers. Userland utilities fetch them
directly, and they were not visible in the SHARED=symlinks case.
Resurrect SHARED=symlinks in Makefile.inc1.
PR: bin/28002
Prodded by: bde
MFC after: 2 weeks
(See commit log for usr.bin/xlint/Makefile,v 1.11 for what was wrong
with enabling build of lint libraries in rev. 1.12.)
This fixes cross-arch compiles (running binaries for a different arch
when generating lint.7 and lint libraries) and cross-branch compiles
(4.x -> 5.0 buildworld should be working again).
to pick up the correct cross-tools (the compiler executables and binutils)
and special linker files (crt*.o). This is now controlled by a single knob,
TOOLS_PREFIX, when building cross-tools.
Fixed regression in Makefile.inc1,v 1.203 (-nostdinc). This clobbered target
architecture's CFLAGS with building host's CPUTYPE setting in /etc/make.conf,
and had a nice but nasty side effect of exposing some (normally hidden) bugs
in system headers.
(Attempt to move the "-nostdinc -I..." part of CFLAGS into the new CINCLUDES
(modeled after a similar CXXINCLUDES) eventually failed because hard-coding
${WORLDTMP}/usr/include to be the first in the include list does not always
work, e.g. lib/libbind.)
Compensate the -nostdinc removal by making cpp(1) built in the cross-tools
stage to not look for <> header files in the building host's /usr/include
(already committed as gnu/usr.bin/cc/cc_tools/freebsd-native.h, revisions
1.10-1.12, STANDARD_INCLUDE_DIR).
: $ /usr/obj/usr/src/i386/usr/bin/cpp -v /dev/null
:
: Before:
:
: #include <...> search starts here:
: /usr/obj/usr/src/i386/usr/include
: /usr/include
: End of search list.
:
: After:
:
: #include <...> search starts here:
: /usr/obj/usr/src/i386/usr/include
: /usr/obj/usr/src/i386/usr/include
(Disabling the use of GCC_INCLUDE_DIR in the FREEBSD_NATIVE case would fix
the duplicate above.)
Get rid of the (now unneeded) -I${DESTDIR}/usr/include magic in bsd.prog.mk
and bsd.lib.mk. Finish the removal of LDDESTDIR in bsd.lib.mk,v 1.55 -- we
no longer have users of it.
The required changes to gcc were already committed as contrib/gcc.295/gcc.c,
revisions 1.23 and 1.24.
Basically, this allows for the changes above plus makes gcc(1) persistent
about path configuration, whether it's configured as a native or a cross
compiler:
: $ /usr/obj/usr/src/i386/usr/bin/cc -print-search-dirs
: install: /usr/obj/usr/src/i386/usr/libexec/(null)
: programs: /usr/obj/usr/src/i386/usr/libexec/elf/:/usr/obj/usr/src/i386/usr/libexec/
: libraries: /usr/obj/usr/src/i386/usr/lib/
:
: $ /usr/obj/alpha/usr/src/i386/usr/bin/cc -print-search-dirs
: install: /usr/obj/alpha/usr/src/i386/usr/libexec/(null)
: programs: /usr/obj/alpha/usr/src/i386/usr/libexec/elf/:/usr/obj/alpha/usr/src/i386/usr/libexec/
: libraries: /usr/obj/alpha/usr/src/i386/usr/lib/
Reviewed by: bde, obrien
Finish it by adding kerberos5/lib/libvers to the build-tools list.
(I didn't notice it before because I tested my fix in a -DNOCLEAN
environment, and static libc.a was already in ${WORLDTMP}/usr/lib,
and libvers's make-print-version build tool used it for linking.)
Spotted by: John Indra <maverick@office.naver.co.id>
(make-roken is a build tool). This bug was hiding itself after
a just fixed bug in cross-linker (binutuils/ld/Makefile,v 1.20).
The bug was fatal for cross builds; for example, an alpha binary
(make-roken) was attempted to be run on i386.
Added make-roken to the list of build-tools in libasn1. It only
worked because another build tool needs make-roken implicitly:
(build-tools: asn1_compile: print_version.o: roken.h: make-roken).
Spotted by: nectar
dependencies in the correct place, record the fact that -lssh
depends on -lcrypto and -lz.
Removed false dependencies on -lz (except ssh(1) and sshd(8)).
Removed false dependencies on -lcrypto and -lutil for scp(1).
Reviewed by: markm
This backs out (sort of) delta 1.18 to perl/miniperl/Makefile.
Update to the ld(1) comment by peter in this revision:
ld(1) built as part of the cross-tools stage of buildworld has
been fixed to look for dynamic dependencies in the right place,
${WORLDTMP}/usr/lib, effective binutils/ld/Makefile,v 1.20.
Approved by: markm
In src/Makefile rev 1.232, environment variable PATH is set explicitly
to /sbin:/bin:/usr/sbin:/usr/bin. As a result, binaries located on
non-standard path cannot be executed without full-path (it's the change
of this revision).
However, cvsup is not in our base system -- you lose if you try to
"make update" without setting SUP in make.conf or command line argument.
I think it is safe to assume that cvsup is located at /usr/local/bin,
and it would help other people who first try to do "make update".
PR: 31932
MFC after: 1 day
Introduce ${TARGET} defaulting to ${MACHINE} which should be set to
whatever your target ${MACHINE} is, and use that with world-related
stages. That is, to build pc98 on alpha, one now needs to set both
TARGET_ARCH=i386 and TARGET=pc98.
The scope of ${TARGET} is limited to Makefile.inc1 and cross-tools.
In particular, this change was tested to fix:
1. Cross building of "alpha" on i386. The breakage was introduced
by rev. 1.10 to sbin/i386/Makefile (missing <machine/cronyx.h>).
2. Descending into machine-specific subdirs for a different arch.
Previously, sbin/i386 and usr.sbin/boot0cfg were descended into
when cross-building "alpha" or pc98 on i386.
3. Fixes pc98 cross-building which was horribly broken, caused by
not setting MACHINE correctly (most ${MACHINE} == pc98 checking
Makefiles put -DPC98 to CFLAGS).
1. To cross-build, one now needs to set TARGET_ARCH, and not the
MACHINE_ARCH. MACHINE_ARCH should never be changed manually!
2. Initialize DESTDIR= explicitly for bootstrap-tools, build-tools,
and cross-tools stages. This fixes broken header and library
dependencies problem. We build them in the host environment,
and obviously want them to depend on host headers and libraries.
The problem with broken header dependencies for bootstrap-tools
and cross-tools was already partially solved (see BOOTSTRAPPING
tests in bsd.prog.mk and bsd.lib.mk), but it was still there for
build-tools if the user ran "make world DESTDIR=/foo". Also,
for all of these stages, the library dependencies were broken
because of how bsd.libnames.mk define DPADD members.
We still provide a glue to install bootstrap- and cross-tools
under the ${WORLDTMP}.
Removed PATH overrides for bootstrap-, build-, and cross-tools
stages. There is just no reason why we would need to override
it, and the hacks to clean up the ${WORLDTMP} in the -DNOCLEAN
case are no longer needed with fixes from this step.
That is, we now never use ${WORLDTMP} headers and libraries,
and we don't use any ${WORLDTMP} installed binaries during
these stages. Again, these stages depend solely on the host
environment, including compiler, headers, and libraries.
3. Moved "miniperl" back from cross-tools (it has nothing to do
with a cross-compiler) to build-tools where it belongs. The
change from step 1 let to do this. Also, to make this work,
build-tools targets of "cc_tools" and "miniperl" were modified
to call "depend". Here follow the detailed explanations.
There are two categories of build tools, for now. In the first
category there are "cc_tools" and "miniperl". They occupy the
whole (sub)directory, and nothing needs to be done in this
subdirectory later during the "all" stage. They are also
constructed using system makefiles. We must build the .depend
early in the build-tools stage because:
1) They use (and depend on) the host environment.
2) If we don't do this in build-tools, the "depend" stage of
buildworld will do this for us; wrong library and header
dependencies will be recorded (DESTDIR=${WORLDTMP}) and,
what's worse, the "all" stage may then clobber the
build-architecture format tools (that we built in the
build-tools stage) with the target-architecture format
ones, breaking cross build.
In the second category there are all other build-tools. They
share their directory with the "main" module that needs them
in the "all" stage, and they don't show up themselves in the
.depend file. The portion of this fix was already committed
in gnu/usr.bin/cc/cc_tools/Makefile,v 1.52.
4. "libperl" is no longer a build tool, and "miniperl" is the
stand-alone application. I had to make this change because
build-tools and "all" stages share the same object directory.
Without this change, if we cross compile, libperl.a is first
built for the build architecture during the build-tools stage
(for the purposes of immediate linkage with "miniperl").
Later on, the "all" stage sees this library as up-to-date,
and doesn't rebuild it. The effect is that the wrong format
static libperl library is installed with installworld.
5. Fixed "includes" to install secure/lib/libtelnet headers if
required.
Reviewed by: bde
1.) prefix all functions in the library with devstat_ (compatability
functions are available for all functions that were chaned in an
incompatible way, but are deprecated).
2.) Add a pointer to a kvm_t as the first argument to functions that
used to get their information via sysctl; they behave the same
as before when NULL is passed as this argument, otherwise, the
information is obtained via libkvm using the supplied handle.
3.) Add a new function, devstat_compute_statistics(), that is intended
to replace the old compute_stats() function. It offers more
statistics data, and has a more flexible interface.
libdevstat does now require libkvm; a library depedency is added, so
that libkvm only needs to be explicitely specified for statically linked
programs.
The library major version number is bumped.
Submitted by: Sergey A. Osokin <osa@freebsd.org.ru>, ken (3)
Reviewed by: ken
want host headers during `buildworld'.
- During `buildworld', install headers in a "copy" mode
until we decide what to do with the (currently broken)
SHARED=symlinks.
- Temporarily run `buildworld' with -DNO_WERROR, which
effectively disabled the -Werror bit of recently added
WARNS=X feature. This is required because adding the
-nostdinc bit back revealed bugs in some header files
that were hiding after not using -nostdinc.
It is unclear currently how exactly (and why) -nostdinc
affects gcc(1) warnings.
file during the bootstrapping process of a buildworld and contained the
last isc include file to be installed. It was meant to be a directory for
the isc include files.
Some of the major changes include:
- The SCSI error handling portion of cam_periph_error() has
been broken out into a number of subfunctions to better
modularize the code that handles the hierarchy of SCSI errors.
As a result, the code is now much easier to read.
- String handling and error printing has been significantly
revamped. We now use sbufs to do string formatting instead
of using printfs (for the kernel) and snprintf/strncat (for
userland) as before.
There is a new catchall error printing routine,
cam_error_print() and its string-based counterpart,
cam_error_string() that allow the kernel and userland
applications to pass in a CCB and have errors printed out
properly, whether or not they're SCSI errors. Among other
things, this helped eliminate a fair amount of duplicate code
in camcontrol.
We now print out more information than before, including
the CAM status and SCSI status and the error recovery action
taken to remedy the problem.
- sbufs are now available in userland, via libsbuf. This
change was necessary since most of the error printing code
is shared between libcam and the kernel.
- A new transfer settings interface is included in this checkin.
This code is #ifdef'ed out, and is primarily intended to aid
discussion with HBA driver authors on the final form the
interface should take. There is example code in the ahc(4)
driver that implements the HBA driver side of the new
interface. The new transfer settings code won't be enabled
until we're ready to switch all HBA drivers over to the new
interface.
src/Makefile.inc1,
lib/Makefile: Add libsbuf. It must be built before libcam,
since libcam uses sbuf routines.
libcam/Makefile: libcam now depends on libsbuf.
libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the
sbuf sources from sys/kern.
bsd.libnames.mk: Add LIBSBUF.
camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically
linked, we can't depend on the dynamic linker
to pull in libsbuf.
camcontrol.c: Use cam_error_print() instead of checking for
CAM_SCSI_STATUS_ERROR on every failed CCB.
sbuf.9: Change the prototypes for sbuf_cat() and
sbuf_cpy() so that the source string is now a
const char *. This is more in line wth the
standard system string functions, and helps
eliminate warnings when dealing with a const
source buffer.
Fix a typo.
cam.c: Add description strings for the various CAM
error status values, as well as routines to
look up those strings.
Add new cam_error_string() and
cam_error_print() routines for userland and
the kernel.
cam.h: Add a new CAM flag, CAM_RETRY_SELTO.
Add enumerated types for the various options
available with cam_error_print() and
cam_error_string().
cam_ccb.h: Add new transfer negotiation structures/types.
Change inq_len in the ccb_getdev structure to
be "reserved". This field has never been
filled in, and will be removed when we next
bump the CAM version.
cam_debug.h: Fix typo.
cam_periph.c: Modularize cam_periph_error(). The SCSI error
handling part of cam_periph_error() is now
in camperiphscsistatuserror() and
camperiphscsisenseerror().
In cam_periph_lock(), increase the reference
count on the periph while we wait for our lock
attempt to succeed so that the periph won't go
away while we're sleeping.
cam_xpt.c: Add new transfer negotiation code. (ifdefed
out)
Add a new function, xpt_path_string(). This
is a string/sbuf analog to xpt_print_path().
scsi_all.c: Revamp string handing and error printing code.
We now use sbufs for much of the string
formatting code. More of that code is shared
between userland the kernel.
scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly
useful in the first place.
Add a new error action, SS_REQSENSE. (Send a
request sense and then retry the command.)
This is useful when the controller hasn't
performed autosense for some reason.
Change the default actions around a bit.
scsi_cd.c,
scsi_da.c,
scsi_pt.c,
scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection
timeouts shouldn't be covered by a sense flag.
scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO.
Get rid of the last vestiges of a read/write
interface.
libkern/bsearch.c,
sys/libkern.h,
conf/files: Add bsearch.c, which is needed for some of the
new table lookup routines.
aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if
CAM_NEW_TRAN_CODE is defined.
sbuf.h,
subr_sbuf.c: Add the appropriate #ifdefs so sbufs can
compile and run in userland.
Change sbuf_printf() to use vsnprintf()
instead of kvprintf(), which is only available
in the kernel.
Change the source string for sbuf_cpy() and
sbuf_cat() to be a const char *.
Add __BEGIN_DECLS and __END_DECLS around
function prototypes since they're now exported
to userland.
kdump/mkioctls: Include stdio.h before cam.h since cam.h now
includes a function with a FILE * argument.
Submitted by: gibbs (mostly)
Reviewed by: jdp, marcel (libsbuf makefile changes)
Reviewed by: des (sbuf changes)
Reviewed by: ken
would have taken care of the possibility of buildkernel crossing over
from one binutils set to another. Back out the part about turning off
'make clean' if the 'make depend' is still active, but add a NO_KERNELCLEAN
target instead that works like NOCLEAN but just for the kernel.
to preceed a buildkernel.
The buildworld is still required when upgrading across major releases,
across binutil upgrades and when config changes version. If
buildkernel breaks, and you haven't done a buildworld, then do not
complain unless you do a buildworld and it still breaks.
already used by the kernel makefiles themselves, and this leads to a lot
of trouble when people put "KERNEL=MYKERNEL" in make.conf. Bite the bullet
and change it to KERNCONF instead, before it gets too far entrenched.
The kernel Makefiles use ${KERNEL} as the name of what to install the
kernel as, eg: /boot/${KERNEL}/kernel or /${KERNEL}. This leads to much
unhappiness with things like /LOCAL instead of /kernel. buildkernel is
severely limited as it is only useful directly after a buildworld.
Reviewed by: jhb
not right because rtermcap would be reading the *host* termcap, not
from the termcap in the src tree. Besides, /usr/sbin/sysinstall
(not the crunched one in /stand) should use the runtime termcap
not the precompiled set.
The distribute target is basicly the same as an install. For
perl, this means that miniperl is needed. Since miniperl is
only present in the object directory, we need to make sure
the path is set correctly. To do this, we have make release
use a new distribworld target that sets the path before doing
a make distribute.
o Move building libperl and miniperl from build-tools to
cross-tools. libperl uses MACHINE_ARCH to determine the
right configuration, which doesn't match the build
machine when cross-building if they are built as build-
tools.
o Since miniperl needs to be built as a cross-tool, it
needs to be installed under /usr/obj so that it can be
used (cross-tools have a special object directory to
avoid build conflicts. As a downside, you can't easily
run cross-tools from their object directory). Remove
the install and distribute override targets. To avoid
having miniperl installed by installworld, remove it
from SUBDIR.
o We can't pickup miniperl from the object directory but
since it's installed, depend on PATH. This is save,
because the makefiles are run with a known path.
o Build libperl again as part of the library target. A
_libperl variable existed, but it was never defined.
o Add chmod to the list of saved tools, because perl
conditionally uses it during install.
The bootstrap-tools and cross-tools targets are modified to
avoid building profiled and shared libraries. While here,
have these targets build static binaries instead of shared
binaries.
Approved by: markm
when using the egcs and gcc-devel ports, along with GCC built from stock
public FSF sources. With out this change, FreeBSD will be removed from
the list of systems GCC 3.0 must be evaluated on before release. With
the effort some of us put into getting FreeBSD on this list, we should
not turn this effort into a waste, else we might not be worth fighting
for in the future. (note that Alpha and IA-64 versions of crt{i,n}.S
are needed)
* Switch from our own crt{begin,in} to those created from GCC's crtstuff.c.
This will allow us to switch to DWARF2 exceptions in the future, along with
staying in sync with any future GCC requirements.
* Break out our ELF branding bits into a seperate file. Currently this
is now included by our crt1.c files (since this functionality was part of
our native crtbegin.c). Later crtbrand.o will be merged in the creation
of crti.o.
If a user decides to forego a make depend during "make buildkernel",
they should get what they deserve if no previous make depend has
been run for that kernel. Instead, the build process includes
special instructions to unconditionally rebuild aicasm. When aicasm
moved to its own directory, this hack broke.
Correct the hack until a get buy off on killing it.
SUPFLAGS when a 'make update' is run. This means that the supfile
doesn't need to be edited because the -h will override the
CHANGE_THIS.FreeBSD.org host.
if their SUPFILE variables are defined _and_ NO_PORTSUPDATE and
NO_DOCUPDATE respectively are not defined.
Previously, only ports was updated and there was no way to prevent
this without undefining its SUPFILE variable.
PR: 17514
Reported by: Udo Erdelhoff <ue@nathan.ruhr.de>
of the buildkernel and installkernel targets where the kernel
was called after the config name.
While here, fix the brokenness of the installkernel target. It
used to use ${IMAKEENV}, but since that has a very restricted
PATH, it couldn't find make(1). Use ${CROSSENV} instead.
binaries we just installed. This allows a future upgrade target to
install a new system without intermediate reboots and also
prevents conflicts for parallel make runs where we might exec a
binary that's being installed at the same time.
