address PR ports/121363 (current day re-opening of PR ports/73797)
to make ports CFLAGS more independent of src/'s CFLAGS WRT aliasing.
Discussed with: brooks
Change the default CFLAGS to match the simple defaults that the
tinderboxes use. By using -fno-strict-aliasing by default we are
choosing to ignore problems in code which had the potential to
shoot ourselves in the foot.
in the default CFLAGS, we're in the middle of a ports freeze, so we can't
really go making the corresponding change to the ports mk files.
I'll take -fno-strict-aliasing out again when the ports freeze ends.
This time, change the default CFLAGS to match the simple defaults that
the tinderboxes use. That is, don't use -fno-strict-aliasing by default.
My last attempt to re-anable -Werror gave me a lesson in what strict
aliasing is all about. There was code in libthr that wasn't 64-bit clean.
The default use of -fno-strict-aliasing hid that.
By using -fno-strict-aliasing by default we were choosing to ignore
problems in code which had the potential to shoot ourselves in the
foot. Sometimes it would be the 64-bit foot. I have both feet. The left
ones are 32 bits and the right ones are 64 bits. Don't ask about my
endian orientation. :-)
The -fno-strict-aliasing compiler arg can still be used if NO_STRING_ALIASING
is define in make.
We are early in the FreeBSD 8 development, so we have the opportunity to
wait and see if this works for us. I am sure that people will complain.
We can easily revert this. All I ask is that we take sides: clean code or
not. YMMV.
Note that by using -fno-strict-aliasing the build won't actually break.
Only where WARNS is set (and -Werror is used) will a compiler warning break
the build. The use of WARNS levels implies (to me at least) that the
developer has taken some care to make the code pass basic checks. This
commit makes those checks just a little bit more strict.
__MAKE_SHELL variable. This feature isn't supposed to be in wide
use, but it's needed now to make `installworld' independent of the
stock binaries and libs so that radical ABI changes can go in safely.
Reviewed by: ru
"There seems to be some continuing discussion about how this is best fixed,
and we'd like to get Alexander (as our gcc guru) to opine on a final
solution before picking one. In the mean time, could you back out the
original commit (sys.mk:1.89)?"
Approved by: re (rwatson)
default. Grepping through src shows only gnu/usr.bin/groff which doesn't
use it in src build and OpenSSH for which this was a NOOP.
Discussed with: des, ru
Approved by: ru
(getting zillions of warnings). Building an old release uses that release's
sys.mk which does not switch on these warnings, so make will be silent.
They can be switch on on the command line with the -x option to make.
This has been tested by building RELENG_5_4 on CURRENT.
by adding WANT_IPFILTER to /etc/make.conf. Note that this is only a partial
hack and only works when building the world and kernel the 'sanctioned' way.
I hope that this hack is only temporary and can be reverted soon.
given -n. For POLA reasons this behaviour is switched on only when
at least two -n flags are given to make. One -n flag keeps the old behaviour
of showing the shell command that would recurse into the sub-directories.
Discussed with: ru
existence of this file. This stopped working more than 4 years ago when
the generation of the error was added in rev.1.44. The .error directive
gives fatal errors, so stale /etc/make.conf.local files must have been
gone away more than 4 years on systems where make(1) works.
CFLAGS in all cases, but POSIX requires a default of -O. Adding
-pipe unconditionally still is still broken for non-gcc compilers
in the non-POSIX case.
libraries to be reported as up-to-date.
Before:
# make -f /dev/null nonexistent.a
`nonexistent.a' is up to date.
After:
# make -f /dev/null nonexistent.a
make: don't know how to make nonexistent.a. Stop
PR: bin/44137 (part of)
to creating the tags file using ctags(1). Defaults to "gtags".
Made GTAGSFLAGS and HTAGSFLAGS overrideable, added CTAGSFLAGS.
Folded bsd.prog.mk version of `tags' into bsd.dep.mk.
PR: bin/42852
o Make it possible to prevent parts of the tree from being linted
(say) during a 'make world' by setting NOLINT in a leaf Makefile.
o Make "make lint" work (better) for executable programs.
o Clean up (nuke!) a syntax damaged pipeline.
under way to move the remnants of the a.out toolchain to ports. As the
comment in src/Makefile said, this stuff is deprecated and one should not
expect this to remain beyond 4.0-REL. It has already lasted WAY beyond
that.
Notable exceptions:
gcc - I have not touched the a.out generation stuff there.
ldd/ldconfig - still have some code to interface with a.out rtld.
old as/ld/etc - I have not removed these yet, pending their move to ports.
some includes - necessary for ldd/ldconfig for now.
Tested on: i386 (extensively), alpha
to use ``.if defined()'' inside bsd.own.mk to test for defines
in individual makefiles. For example, setting DEBUG_FLAGS in
Makefile didn't take the desired effect on the STRIP assignment.
Added bsd.init.mk (like in NetBSD) that handles the inclusion
of ../Makefile.inc and bsd.own.mk from all bsd.*.mk files that
"build something".
Back out bsd.own.mk,v 1.15: moved OBJFORMAT initialization back
to sys.mk (several source tree makefiles want to check it early)
and removed MACHINE_ARCH initialization (it's hard to see from
looking at the commitlogs what the problem was at the time, but
now it serves no purpose).
Prohibit the direct inclusion of bsd.man.mk and bsd.libnames.mk.
Protect bsd.obj.mk from repetitive inclusion. Prohibiting the
direct inclusion of bsd.obj.mk might be a good idea too.
(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).
Use this where we are now using /etc/make.conf.
This allows people to override the current default of always including
/etc/make.conf. Setting __MAKE_CONF to /dev/null disables it
completely, while setting it to something else allows one to override
what is on the system. This can be desirable in situations where a
machine has many users and some of them want different defaults, or
defaults appropriate to cross building to be different than those for
normal building.
Not objected to by: arch@
lint, so this is turned off by default. Setting WANT_LINT will turn
on generation of lint libraries for /usr/libdata/lint/*.ln.
Reviewd by: silence in -audit.
supposed to be edited by the user and didn't define important things,
thus we can just skip it (that's where it differs from the make.conf.local
change).
Submitted by: ru
discussed on the arch@ mailinglist (after repo-copy).
sys.mk will .error if it finds /etc/defaults/make.conf but include
it anyways (this is the same behaviour as with the make.conf.local
removal).
/usr/share/examples/etc/make.conf has BDEFLAGS commented out now,
since it's only an example file.
Adjust all textes that talk about make.conf or defaults/make.conf to
match the new situation.
(While there, I also moved the single suffix C rules beside the double
suffix ones so they are easier to find)
PR: 24438
Submitted by: Georg-W. Koltermann <gwk@sgi.com>
* Rip out MACHINE_CPU stuff from sys.mk and include a new <bsd.cpu.mk>
after we pull in /etc/make.conf. We need to do it afterwards so we can
react to the user setting of the:
* CPUTYPE variable, which contains the CPU type which the user wants to
optimize for. For example, if you want your binaries to only run on an
i686-class machine (or higher), set this to i686. If you want to support
running binaries on a variety of CPU generations, set this to the lowest
common denominator. Supported values are listed in make.conf.
* bsd.cpu.mk does the expansion of CPUTYPE into MACHINE_CPU using the
(hopefully) correct unordered list of CPU types which should be used on
that CPU. For example, an AMD k6 CPU wants any of the following:
k6 k5 i586 i486 i386
This is still an unordered list so the client makefile logic is simple -
client makefiles need to test for the various elements of the set in
decreasing order of priority using ${MACHINE_CPU:M<foo>}, as before.
The various MACHINE_CPU lists are believed to be correct, but should be
checked.
* If NO_CPU_CFLAGS is not defined, add relevant gcc compiler optimization
settings by default (e.g. -karch=k6 for CPUTYPE=k6, etc). Release
builders and developers of third-party software need to make sure not to
enable CPU-specific optimization when generating code intended to be
portable. We probably need to move to an /etc/world.conf to allow the
optimization stuff to be applied separately to world/kernel and external
compilations, but it's not any worse a problem than it was before.
* Add coverage for the ia64/itanium MACHINE_ARCH/CPUTYPE.
* Add CPUTYPE support for all of the CPU types supported by FreeBSD and gcc
(only i386, alpha and ia64 first, since those are the minimally-working
ports. Other architecture porters, please feel free to add the relevant
gunk for your platform).
Reviewed by: jhb, obrien
set the variable until you rebuild it, and the alternative is to be stuck
playing games with ``.if defined(MACHINE_CPU) && ... '' for all eternity.
We now set up the reasonable default for i386 and alpha here -- given this
it probably makes sense to remove the corresponding code from make(1).
through the use of a new build directive, MACHINE_CPU, which contains a
list of the CPU generations/features for which optimizations are desired.
This feature will be extended to cover the ports tree in the future.
Currently OpenSSL provides optimizations for i386, i586 and i686-class
CPUs. Currently it has not been tested on an i386 or i486.
Teach make(1) to provide sensible defaults for MACHINE_CPU if it is not
defined (namely, the lowest common denominator CPU we support for each
architecture). Currently this is i386 for the i386 architecture and ev4
for the alpha. sys.mk also sets the variable as a last resort for
consistency with MACHINE_ARCH and bootstrapping from very old versions of
make.
Benchmarks show a significant speed increase even in the i386 case, with
additional improvements for i586 and i686 systems. For maximum performance
define MACHINE_CPU=i686 i586 i386 in /etc/make.conf.
Based on a patch submitted by: Mike Silbersack <silby@silby.com>
Reviewed by: current
recently added definitions from sys.mk to bsd.own.mk. Include the
src-relative bsd.own.mk in src/Makefile to pick up all new definitions.
Don't check that MACHINE_ARCH is defined in src/Makefile, since it is
(and should have been) guaranteed to be defined.
Notes:
- We no longer use -fgnu-runtime in bsd.lib.mk, since it is the default
and bsd.lib.mk is the wrong place to override it.
- Gnu C doesn't have a special compiler driver for Objective C like it
does for C++. The defaults are suitable for Gnu C. Use `OBJCLIBS='
in /etc/make.conf for POC.
it is built with this defined (which it isn't by default). This change
to sys.mk treats the absence of MACHINE_ARCH as i386 on the assumption
that it will be appropriately defined (as something else) on any other
architecture. When building FreeBSD's make with NetBSD tools, both
MACHINE and MACHINE_ARCH are correctly set (e.g. when bootstrapping
FreeBSD's make on NetBSD/mvme68k, MACHINE=mvme68k and
MACHINE_ARCH=m68k). This isn't really needed for the alpha which
has both defined as 'alpha', but I thought it was worth getting the
distinction between a MACHINE and a MACHINE_ARCH correct now.
Now, shouldn't PC98 have MACHINE=pc98 and MACHINE_ARCH=i386 ??!!
.if in Makefiles. bsd.prog.mk and bsd.lib.mk do not depend on it however.
Allow overriding of the -soname arg when building the lib*crypt.so* libs
since libdescrypt.so and libscrupt.so both need a -soname of libcrypt.so
so that the symlink is obeyed at runtime rather than at compile time.
since 2.1.x make(1) apparently does not have the -m switch to set both
the the bsd.*.mk and sys.mk location, and this breaks 'make world' from a
2.1.x system.
note, using "-Wl,-f" to generate a library objects list doesn't work
anymore since the hack to ld hasn't been incorporated into binutils-2.8.
(and the -f switch is used for something else already)
This is disabled by default, don't panic! :-)
This will make a number of things easier in the future, as well as (finally!)
avoiding the Id-smashing problem which has plagued developers for so long.
Boy, I'm glad we're not using sup anymore. This update would have been
insane otherwise.
be easy to change to /usr/X11R6 if and when the time comes. This is
to deal with things like xditview which otherwise had hardcoded assumptions
about where X lived. Yuck.
Submitted by: jkh
previous commit:
+ Everything is initialized using ?= instead of =.
+ Nicer formatting (more white space).
+ .c: rule.
Add macros ECHO and ECHODIR. Both are normally `echo', but when
the make flags include -s, ECHO is set to `true' and when the make
flags include two or more -s's ECHODIR is set to `true'. @${ECHO}
should be used instead of @echo in most cases. ${ECHODIR} is
intended to be used mainly for messages about directory names.
