That way the files are correctly taggued for pkgbase
Reviewed by: bapt, emaste (both earlier version)
Differential Revision: https://reviews.freebsd.org/D29171
MFC after: 2 weeks
As suggested in D27598. This also supports MK_WERROR.clang=no and
MK_WERROR.gcc=no to support the existing NO_WERROR.<compiler> uses.
Reviewed By: brooks
Differential Revision: https://reviews.freebsd.org/D27601
We are seeing regular build failures due to libc.so being installed again and
another parallel make job tries to read the partially written libc.so at the
same time. When building with -j32 or higher this almost always happens on
the first clean build (subsequent incremental builds always work fine).
Using -S should "fix" the "section header table goes past the end of the
file: e_shoff = 0x..." errors that have started to plague our builds.
We originally thought this only affected CheriBSD, but I just got the same
error while building the latest upstream FreeBSD.
The real fix should be to not install libraries twice, but until then this
workaround is needed.
Original patch by jrtc27@, I only made some minor changes to the comment.
Obtained from: CheriBSD (49837edd3e)
Reviewed By: markj, bdrewery
Differential Revision: https://reviews.freebsd.org/D27102
There are two options:
- WITH_INIT_ALL_ZERO: Zero all variables on the stack.
- WITH_INIT_ALL_PATTERN: Initialize variables with well-defined patterns.
The exact pattern are a compiler implementation detail and vary by type.
They are somewhat documented in the LLVM commit message:
https://reviews.llvm.org/rL349442
I've used WITH_INIT_ALL_* to match Microsoft's InitAll feature rather
than naming them after the LLVM specific compiler flags.
In a range of consumer products, options like these are used in
both debug and production builds with debugs builds using patterns
(intended to provoke crashes on use of uninitialized values) and
production using zeros (deemed more likely to lead to harmless
misbehavior or NULL-pointer dereferences).
Reviewed by: emaste
Obtained from: CheriBSD
Sponsored by: DARPA
Differential Revision: https://reviews.freebsd.org/D27131
It appears this was changed from ln to use install in rS245752. I noticed
this because my buildenv was setting INSTALL=install -U -M //METALOG
and then these links fail to be created with the following error:
install: open //METALOG: Permission denied
Reviewed By: brooks
Differential Revision: https://reviews.freebsd.org/D26618
As we do for shared library binaries, pass -S to install(1) when
installing symlinks. Doing so helps avoid transient failures when
libraries are being reinstalled, which seems to be the root cause of
spurious libgcc_s.so link failures during CI builds.
PR: 233769
Reviewed by: emaste
MFC after: 1 week
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D26453
Use of ranlib or lorder is no longer necessary with current linkers
(probably anything newer than ~1990) and ar's ability to create an object
index and symbol table in the archive.
Currently the build system uses lorder+tsort to sort the .o files in
dependency order so that a single-pass linker can use them. However,
we can use the -s flag to ar to add an index to the .a file which makes
lorder unnecessary.
Running ar -s is equivalent to running ranlib afterwards, so we can also
skip the ranlib invocation.
Similarly, we don't have to pass the .o files for shared libraries in
dependency order since both ld.bfd and ld.lld will correctly resolve
references between the .o files.
This removes many fork()+execve calls for each library so should speed up
builds a bit. Additionally lorder.sh uses a regular expression that is not
supported by the macOS libc or glibc and results in many warnings when
cross-building (see D25989).
There is one functional change: lorder.sh removed duplicated .o files
from the linker command line which now no longer happens. I fixed the duplicates
in the base system in r364649. I also checked the ports tree for uses of
bsd.lib.mk and found one duplicate source file which I fixed in r548168.
Most ports use CMake/autotools rather than bsd.lib.mk but if this breaks any
ports that I missed in my search please let me know.
Avoiding the shell script actually speeds up the linking step noticeably: I
measured how long it takes to rebuild the .a and .so files for lib/libc using a
basic benchmark: `rm $LIBC_OBJDIR/*.so* $LIBC_OBJDIR/*.a* && /usr/bin/time make -DWITHOUT_TESTS -s > /dev/null`
Without this change ~4.5 seconds and afterwards ~3.1 seconds.
Looking at truss -cf output we can see that the number fork() system
calls goes down from 27 to 12 (and the speedup while tracing is more
noticeable: 81 seconds -> 65 seconds).
See also https://www.gnu.org/software/coreutils/manual/html_node/tsort-background.html
for some more background:
This whole procedure has been obsolete since about 1980, because Unix
archives now contain a symbol table (traditionally built by ranlib, now
generally built by ar itself), and the Unix linker uses the symbol table
to effectively make multiple passes over an archive file.
Or alternatively https://www.unix.com/man-page/osf1/1/lorder/:
The lorder command is essentially obsolete. Use the following command in
its place: % ar -ts file.a
Reviewed By: emaste, imp, dim
Differential Revision: https://reviews.freebsd.org/D26044
-development is long and awkward, and is also inconsistent with prior art
from the Linux world, which uses -dev (Debian) or -devel (Red Hat). Follow
the Debian convention, and similarly for debug info packages.
Also remove redundant pkgbase development tag from includes. We already tag
include files with package=runtime,dev; there is no need to separately tag
them as dev.
Discussed with: bapt
Reviewed by: manu
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D24139
This option was added as a transition aide when symbol versioning was
first added. It was enabled by default in 2007 and is supported even
by the old GPLv2 binutils. Trying to disable it currently fails to
build in libc and at this point it isn't worth fixing the build.
Reported by: Michael Dexter
Reviewed by: imp
Differential Revision: https://reviews.freebsd.org/D24637
Original commit message:
bsd.lib.mk: Do not include bsd.incs.mk for INTERNALLIB
f we're building an internal lib do not bother including bsd.incs.mk so we
will not install the headers.
This also "solves" a problem with pkgbase where a libXXX-development package
is created and due to how packages are created we add a dependency to a
libXXX package that doesn't exists.
If we're building an internal lib do not bother including bsd.incs.mk so we
will not install the headers.
This also "solves" a problem with pkgbase where a libXXX-development package
is created and due to how packages are created we add a dependency to a
libXXX package that doesn't exists.
Reported by: pizzamig
Reviewed by: pizzamig bapt emaste
Differential Revision: https://reviews.freebsd.org/D24166
Profiling library archives are part of the development environment; they
don't need to be in separate -profile packages.
(In fact we can probably just eliminate the _p.a archives assuming that
profiling will be done using hwpmc etc., but that is a change for later.)
Discussed with: bapt, manu
Sponsored by: The FreeBSD Foundation
arichardson has an actual fix for the same issue that this was working
around; given that we don't build with llvm today, go ahead and revert the
workaround in advance.
bsd.cpu.mk is included by bsd.init.mk before bsd.linker.mk, so it
was always setting the flag since LINKER_FEATURES wasn't defined.
Reported by: mhorne
Reviewed by: imp, mhorne
Sponsored by: DARPA
Differential Revision: https://reviews.freebsd.org/D23076
Diff partially stolen from CheriBSD; these bits need -Wl,-z,notext in order
to build in an LLVM world. They are needed for all flavors/sizes of MIPS.
This will eventually get fixed in LLVM, but it's unclear when.
Reported by: arichardson, emaste
Differential Revision: https://reviews.freebsd.org/D21696
GCC uses "dynamic" TLS models when -fpic or -fPIC is explicitly
specified on the command line (which is only true for shared libraries).
It uses "static" (or "exec") TLS models otherwise. In particular, GCC
does _not_ use dynamic TLS models when PIC is implicitly enabled (which
it is on MIPS), only if a PIC flag is explicitly provided.
llvm uses "dynamic" TLS models if PIC is enabled either via a PIC flag
or if it is implicily enabled (as on MIPS64). This means that llvm on
MIPS64 always uses "dynamic" TLS models. However, dynamic TLS models
do not work for static binaries and libraries as the __tls_get_addr
function they invoke is only defined in rtld.
Written by: jhb
Reviewed by: arichardson
Differential Revision: https://reviews.freebsd.org/D21699
The default package use to be FreeBSD-runtime but it should only contain
binaries and libs enough to boot to single user and repair the system, it
is also very handy to have a package that can be tranform to a small mfsroot.
So create a new package named FreeBSD-utilities and make it the default one.
Also move a few binaries and lib into this package when it make sense.
Reviewed by: bapt, gjb
Differential Revision: https://reviews.freebsd.org/D21506
Building binaries as PIE allows the executable itself to be loaded at a
random address when ASLR is enabled (not just its shared libraries).
With this change PIE objects have a .pieo extension and INTERNALLIB
libraries libXXX_pie.a.
MK_PIE is disabled for some kerberos5 tools, Clang, and Subversion, as
they explicitly reference .a libraries in their Makefiles. These can
be addressed on an individual basis later. MK_PIE is also disabled for
rtld-elf because it is already position-independent using bespoke
Makefile rules.
Currently only dynamically linked binaries will be built as PIE.
Discussed with: dim
Reviewed by: kib
MFC after: 1 month
Relnotes: Yes
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D18423
The linker's -z now flag sets the DF_BIND_NOW flag, which signals to the
runtime loader that all relocation processing should be performed at
process startup rather than on demand. In combination with lld's
default of enabling relro this causes the GOT to be made read-only when
the process starts, preventing straightforward GOT overwrite attacks.
Shawn Webb discovered a failure on HardenedBSD with BIND_NOW and ifunc
use, which resulted in my rtld fix in r340137. Add a BIND_NOW knob as
it is trivial to do so and is a useful ELF hardening feature. This
change is equivalent to HardenedBSD's but not identical as there are
other diffs/conflicts nearby.
Note that our ELF Tool Chain readelf does not currently decode the
DF_BIND_NOW flag - see PR232983.
Reviewed by: brooks
MFC after: 1 week
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D17846
This paves the way for moving config files out of head/etc and into the
directories with the src.
Approved by: bapt (mentor)
Differential Revision: https://reviews.freebsd.org/D16406
Rtld is not compatible with SSP, and since we link libc_pic.a to rtld
to have the basic support like memory and string copy functions, we
have to both carefully limit libc use, and to provide the ssp support
shims. This change makes the libc use in rtld more straighforward but
still limited, and allows to remove the shims, to be done in the next
commit.
Submitted by: Luis Pires
Reviewed by: bdrewery, brooks
Differential revision: https://reviews.freebsd.org/D15283
Now that OBJS has grown an OBJS_SRCS_FILTER variable, use this variable
in the computation of BCOBJS and LLOBJS too. Also move BCOBJS and LLOBJS
computation to be next to the OBJS computation: this should both make
the parallel structure clearer and serve to remind people changing OBJS
that parallel changes are required in BCOBJS and LLOBJS.
A side effect of this change is that BCOBJS and LLOBJS will be available
even when LLVM_LINK has not been defined, but that seems like a positive
change: there's no reason we can't ask "what bitcode files would you
generate" just because we can't link those files together into a
complete bitcode representation of a binary or library.
Reviewed by: sjg
Sponsored by: DARPA, AFRL
Differential Revision: https://reviews.freebsd.org/D12701
We previously taught the build system how to create files like libfoo.bc,
but neglected to teach it about cleaning such files up. Rectify this now.
MFC after: 1 week
Sponsored by: DARPA, AFRL
Some makefiles do reachover builds.
In some cases it is convenient to list subdirs of the distribution
in SRCS.
It is not very convenient, or always even desirable to have corresponding
subdirs in .OBJDIR, so OBJS_SRCS_FILTER allows the makefile to choose.
The default value 'R' matches existing practice.
But a makefile can set OBJS_SRCS_FILTER= T (the R gets added by
bsd.init.mk) to avoid the need for subdirs in .OBJDIR
Differential Revision: https://reviews.freebsd.org/D12218
Reviewed by: bdrewery
The -S flag is currently ignored for builds since we filter through
tools/install.sh that is intended for both non-root and cross-builds.
Sponsored by: Dell EMC Isilon
X-MFC-With: r322565
- Include debug symbols in static libraries. This permits binaries
to include debug symbols for functions obtained from static libraries.
- Permit the C/C++ compiler flags added for MK_DEBUG_FILES to be
overridden by setting DEBUG_FILES_CFLAGS. Use this to limit the debug
information for llvm libraries and binaries.
Reviewed by: emaste
Sponsored by: DARPA / AFRL
Differential Revision: https://reviews.freebsd.org/D12025
Also disable this if NO_SAFE_LIBINSTALL is defined.
There is little harm in always using -S and it fixes several issues:
- A race during 'make libraries' where, for example, libgcc_s is being
installed while another library is trying to link against it. This is
possible because libgcc_s is connected in both _prereq_libs and
_startup_libs. The first build (_prereq_libs) sets MK_PROFILE=no
while the 2nd pass (_startup_libs) enables MK_PROFILE. Thus the
libgcc_s library *is* present in WORLDTMP for other libraries to
link to, so serializing further items in _startup_libs is not
required. Just ensuring that libgcc_s is installed atomically (via
rename(2)) is enough. [1]
- Installation to a running system where some library that cannot be
detected, copied and used from the temporary INSTALLTMP with LD_LIBRARY_PATH
that the build itself uses for installation. Such an example is having the
install an NSS module for user lookups that install(1) uses while
concurrently installing the module in another process. This is not
a problem for the FreeBSD base build but can be for downstream
vendors. While this is a very specific case, installation to a
running system with non-atomic library installation is prone to many
problems. A further step still is to install in proper dependency
ordering.
Reported by: dhw many times [1]
Sponsored by: Dell EMC Isilon
MFC after: 2 weeks
== Rationale ==
r295380 introduced "make check" and consolidated means for running
test code in an attempt to simplify running tests. One could either
install files/libraries/programs and run "make check", or run "make check"
with an explicit CHECKDIR, e.g., `make check CHECKDIR=$(make -V.OBJDIR)``.
One criticism that was received is that "make check" should be run with
the intent of making dev->test->commit easier, which means that the target
audience's workflow should be developers. One developer pattern available
in other opensource projects is to run test code from a developer sandbox,
instead of installing to a system.
== Method ==
This approach is slightly different from the standard approach, in the sense
that it builds and installs into a deterministic directory under .OBJDIR (as I call it,
the "sandbox"), then runs "make check" against that. In the event the test
run is successful, the deterministic directory is removed to save space.
== Approach ==
bsd.lib.mk, bsd.prog.mk:
To support this functionality, a new variable `HAS_TESTS` is being added.
HAS_TESTS enables appropriate behavior with bsd.lib.mk and bsd.prog.mk, as
follows:
- Add "make check" as an available target from the directory.
- Pass down appropriate variables via ${TESTS_ENV}, i.e.,
${TESTS_LD_LIBRARY_PATH} and ${TESTS_PATH}.
One should add "HAS_TESTS" to directories containing tests in them, e.g. from
bin/sh/Makefile,
HAS_TESTS=
SUBDIR.${MK_TESTS}+= tests
HAS_TESTS doesn't automatically add the tests subdirectory for flexibility
reasons.
bsd.opts.mk, src.opts.mk:
- The knob ${MK_MAKE_CHECK_USE_SANDBOX} has been added, both to explicitly
direct (internally) when to set a deterministic ${DESTDIR} and to also allow
users to disable this behavior globally, i.e., via src.conf.
- MK_TESTS has been promoted from src.opts.mk to bsd.opts.mk to leverage
syntactic sugar for having MK_TESTS be a dependency for
MK_MAKE_CHECK_USE_SANDBOX, but to also ensure that src.opts.mk isn't required
to use suite.test.mk (which is a dependency of bsd.test.mk).
suite.test.mk:
- beforecheck behavior (when MK_MAKE_CHECK_USE_SANDBOX is enabled) is modified
from a no-op to:
-- Build.
-- Run "make hierarchy" on the sandbox dir.
-- Install the tests/files to the sandbox dir.
- aftercheck behavior (when MK_MAKE_CHECK_USE_SANDBOX is enabled) is modified
from a no-op to:
-- Remove the sandbox dir.
Again, because the dependency order set in bsd.test.mk is
beforecheck -> check -> aftercheck, "make check" will not be run unless
"beforecheck" completes successfully, and "aftercheck" will not be run unless
"beforecheck" and "check" complete successfully.
== Caveats ==
- This target must either be run with MK_INSTALL_AS_USER or as root. Otherwise
it will fail when running "make install" as the default user/group for many
makefiles when calling INSTALL is root/wheel.
- This target must be run from a suitable top-level directory. For example,
running tests from `tests/sys/fs/tmpfs` won't work, but `tests/sys/fs` will,
because `tests/sys/fs/tmpfs` relies on files installed by `tests/sys/fs`.
- Running MK_INSTALL_AS_USER may introduce determinism issues. However, using
it could identify deficiences in tests in terms of needing to be run as
root, which are not properly articulated in the test requirements.
- The doesn't negate the need for running "make installworld" and
"make checkworld", etc. Again, this just is intended to simplify the
dev->test->commit workflow.
== Cleanup done ==
- CHECKDIR is removed; one can use "MK_MAKE_CHECK_USE_SANDBOX=no" to enable
"legacy" (r295380) behavior.
MFC after: 2 months
Relnotes: yes (CHECKDIR removed; "make check" behavior changed)
Requested by: jhb
Reviewed by: arch (silence), testing (silence)
Differential Revision: D11905
Running `make libfoo.ll` or `make libfoo.bc` within a library directory
will now give us an LLVM IR version of the library, and `make foo.full.ll`
or `make foo.full.bc` will give us an IR version of a binary.
As part of this change, we add an LLVM_LINK variable to sys.mk that can be
specified/overridden using an external toolchain.
Reviewed by: bdrewery, brooks
Approved by: rwatson (mentor)
Sponsored by: DARPA, AFRL
Differential Revision: https://reviews.freebsd.org/D8388
Refactor make suffix rules into separate files (one for POSIX and one not),
and rationalise the rules so that bsd.lib.mk can contain only those rules
that are library-specific (.c.po and .c.pico).
This can be accomplished by adding ${STATIC_CFLAGS} to the .c.o rule
unconditionally. STATIC_CFLAGS are only defined for use by sys.mk rules in
lib/libpam/Makefile.inc (see r227797), so it should be safe to include
them unconditionally in sys.mk's .c.o rule (tested by make universe and a
ports exp-run).
Reviewed by: bdrewery, sjg
Approved by: rwatson (mentor)
Sponsored by: DARPA, AFRL
Differential Revision: https://reviews.freebsd.org/D6805
build can break when different source files create the same target
files (case-insensitivity speaking). This is the case for object
files compiled with -fpic and shared libraries. The former uses
an extension of ".So", and the latter an extension ".so". Rename
shared object files from *.So to *.pico to match what NetBSD does.
See also r305855
MFC after: 1 month
Sponsored by: Bracket Computing
Differential Revision: https://reviews.freebsd.org/D7906
This is important to allow a Makefile to override OBJS_DEPEND_GUESS for
handling in META_MODE when its depend files are missing.
Approved by: re (implicit)
Sponsored by: EMC / Isilon Storage Division
A simplified example of the library targets with WITH_DEBUG_FILES is:
libgeom.so.5: libgeom.so.5.full
cp libgeom.so.5.full libgeom.so.5
libgeom.so.5.full:
ln -s libgeom.so.5 libgeom.so
cc -o libgeom.so.5.full *.o
Before, or without, WITH_DEBUG_FILES it is:
libgeom.so.5:
ln -s libgeom.so.5 libgeom.so
cc -o libgeom.so.5 *.o
The problem is that bmake considers the link source for the libgeom.so
link in the libgeom.so.5.full target as being a dependency for
libgeom.so.5.full. That resolves to libgeom.so.5. Thus a cyclic
dependency is created. The result of this is that if libgeom.so.5 is
created with a newer timestamp than libgeom.so.5.full, then
libgeom.so.5.full will be rebuilt on the next build. This causes a
chain reaction of everything in the build relinking, or hitting the
problem itself.
Moving the link creation to the target that actually creates
libgeom.so.5 fixes the problem. The simplest fix here is to just
duplicate the logic.
Submitted by: sjg
Approved by: re (implicit)
This breaks cross-building with WITH_META_MODE since it will rebuild
'build-tools' during the 'everything' phase.
A more proper fix is coming to bmake to implicitly require .META unless
.NOMETA (and other restrictions) are in place.
Adding .META to targets-to-build will ensure that they will rebuild if there
is no .meta file.
Adding it to all SUFFIXES and objects ensures that at least objects will
rebuild if there is no .meta file.
This will be reverted if bmake's behavior changes to rebuild on missing .meta
files.
Sponsored by: EMC / Isilon Storage Division
We want to avoid .text relocations in shared objects. libcrypto was the
only consumer and it is now fixed (as of r299389). Remove the now-unused
support for turning off the linker warning.
Reviewed by: kib
Differential Revision: https://reviews.freebsd.org/D6323
