The boost and jack2 packages fail to build when PARALLEL_JOBS is empty
so instead of using an empty PARALLEL_JOBS don't use it in the MAKE
variable when top-level parallel make is being used.
To simplify the use of top-level parallel make, check the MAKEFLAGS
variable to know automatically if the -j option is being used, also use
the "=" operator instead of the ":=" operator because the MAKEFLAGS
variable can be checked only in a "recursively expanded variable".
The "override" keyword must be used in order to change the automatic
variable "MAKE".
When the top-parallel make is being used the sub-make are called without
specifying the "-j" option in order to let GNU make share the job slots
specified in the top make. This is done because GNU make is able
to share the job slots available between each instance of make so if you
want to increase the number of jobs you just need to increase the <jobs>
value in the top make -j<jobs> command.
If we specify the -j<jobs> option in each instance of make, it is less
efficient, e.g. in a processor with 8 cores we specify -j9 in each instance:
the number of processes goes up to 81 because each sub-make can execute
9 processes. The excessive number of processes is not a good thing
because in my tests even -j16 is slower than -j9.
Instead if we don't specify the -j<jobs> option in the sub-make, the top
make share the job slots automatically between each instance, so the
number of process in this examples goes up to 9 that is faster than
using up to 81 processes.
e.g. when the -j3 option is specified only in the top make:
possible state n. 1:
process 1 - <packagea>-build
process 2 - <packagea>-build
process 3 - <packagea>-build
possible state n. 2:
process 1 - <packagea>-extract
process 2 - <packageb>-configure
process 3 - <packagec>-build
possible state n. 3:
process 1 - <packagea>-build make -j1
process 2 - <packageb>-build make -j1
process 3 - <packagec>-build make -j1
Signed-off-by: Fabio Porcedda <fabio.porcedda@gmail.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
CMake verbose mode is based on VERBOSE environment variable.
* If VERBOSE is exported but empty, only "Dependee ... is newer than
depender ..." messages are shown.
* If VERBOSE is exported and set (whatever the value), all compilation
commands are shown.
VERBOSE is currently systematically exported by Buildroot, even if it
is empty, in the root Makefile, which implies that the "light" verbose
mode - with "Dependee ... is newer than depender ..." messages - is
always enabled.
VERBOSE should only be exported when V=1, which is the standard way to
enable verbose mode in Buildroot.
Signed-off-by: Cédric Marie <cedric.marie@openmailbox.org>
Reviewed-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
BR2_DEFCONFIG should not be present in saved defconfig file.
The use case is:
make qemu_arm_versatile
make savedefconfig BR2_DEFCONFIG=my_custom_defconfig
BR2_DEFCONFIG is set in my_custom_defconfig with an absolute path
to qemu_arm_versatile (value present in .config) and set in
my_custom_defconfig as it is different from default mentioned in
config.in (default is BR2_DEFCONFIG from environment).
On savedefconfig recipe, simply remove BR2_DEFCONFIG from generated file
[Peter: fixup typos and use SED as noted by Arnout]
Signed-off-by: Herve Codina <Herve.CODINA@celad.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
It is possible to end up with a path containing spaces if the kernel
localversion contains spaces.
Be it good practice or not, there are third party vendors which
distribute kernel configuration files for reference platforms which have
quoted strings containing whitespaces in the localversion.
There was already a fix to handle paths with whitespaces or other
special characters when running strip, which consists of using the find
-print0 and xargs -0 pair of arguments, but the kernel module stripping
wasn't included in the fix.
This commit includes the same fix to the kernel module stripping line.
Signed-off-by: Erico Nunes <erico.nunes@datacom.ind.br>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
With this commit, one can now execute the source-check, external-deps
and legal-info targets regardless of the checks normally being done by
packages on the configuration.
Note that we intentionally do not go down the road of adding %-source,
%-legal-info, and the miryad of other targets that could work in such
situations. We only whitelist a few targets that are really useful to
have as nobuild_targets.
[Thomas: also add 'clean' and 'distclean' to the nobuild_targets, as
suggested by Yann.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Some packages do some sanity checks on their configuration, for
example linux checks that the defconfig string is not empty when a
defconfig is used. Such checks are currently always performed, except
when the 'source' target is part of make goals.
This is problematic for two reasons:
- Other targets such as 'source-check', 'external-deps' or
'legal-info', that do not consist in doing a build, cannot be
executed in such situations.
- The current code removes the check as soon as one of the targets is
source. But if there are other non-source targets called at the
same time, the checks are ignored.
This commit therefore introduces an internal variable called
BR_BUILDING, which tells packages if we are actually building or
not. A variable nobuild_targets indicates the targets that we do not
consider as being build targets.
For the moment, nobuild_targets only contains 'source', to be
completely iso-functional.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Now that all the external-deps, source-check and source targets are
properly implemented based on the package infrastructure, the
PACKAGES_SOURCE, TARGET_HOST_DEPS, HOST_DEPS and HOST_SOURCE variables
are no longer needed. This is a good thing since they were anyway
incorrect, as they were only doing a two level recursion in the
dependencies of host packages.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Tested-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
[tested with a randpackageconfig]
Now that all the bits are in place, switch the global 'source' target
to use the package infrastructure logic.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Tested-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
[with 'make source' (actually together with the next patch).]
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
[tested with a randpackageconfig]
This commit switches the implementation of the global source-check
target to use a package infrastructure based mechanism, using the
$(1)-all-source-check target added in the previous commit.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Reviewed-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
make source-check is here to check whether the remote sources for the
current selection of packages are still available. In its current
implementation, since it simply calls recursively a sub-make with the
source target, it can be a noconfig_targets. However, a follow-up
change will make source-check not use a sub-make, which will require
it to no longer be a noconfig_targets.
Therefore, as a preparation, this commit moves source-check outside of
noconfig_targets.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Reviewed-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
This commit changes the global 'external-deps' target to use the newly
introduced per-package <pkg>-all-external-deps, instead of relying on
the 'source' target with a custom DL_MODE.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
When listing defconfig files was moved to its own make target, it was
not added to the general help text.
However, this is a very important topic, so list it.
[Peter: drop ':' character]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Arnout Vandecappelle <arnout@mind.be>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
On some systems (e.g. Fedora 20), the build breaks when parsing include
directives in our /etc/ld.so.conf, with error messages as thus:
/sbin/ldconfig: need absolute file name for configuration file
when using -r
So, enforce the path to the ld.so.conf file to point to our own, in the
target/ directory.
Reported-by: Al West <al.west@v-nova.com>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The first dependency of these targets is
$(BUILD_DIR)/buildroot-config/*conf
so the $(BUILD_DIR)/buildroot-config directory certainly exists.
Reported-by: Robert P. J. Day <rpjday@crashcourse.ca>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
show-targets is only used currently by the graph-depends script, which
already recurses into the dependencies of the selected packages to
build the dependency graph. Therefore, dumping the contents of
$(PACKAGES) and $(ROOTFS_TARGETS) is sufficient: $(HOST_DEPS) and
$(TARGET_HOST_DEPS) will contain packages that are dependencies of
packages already listed in $(PACKAGES), which graph-depends will
discover by itself.
This allows to remove one more usage of $(HOST_DEPS) and
$(TARGET_HOST_DEPS), which is one more step towards their removal.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
This commit changes the implementation of the global 'legal-info'
target to use the newly introduced per-package <pkg>-all-legal-info
target. This allows to avoid using the $(TARGET_HOST_DEPS) and
$(HOST_DEPS) variables that we are trying to remove.
It is worth mentionning that this commit might change the output of
'make legal-info' by making it more correct than it was. With the
existing implementations, we could be missing packages if they were
host packages, or target packages not properly selected in terms of
Config.in dependencies, and with a more than a two-level deep
dependency from a target package properly selected at the Config.in
level. This is because our previous logic was simply taking all
packages in the "TARGETS" (now called "PACKAGES") variable, which are
only the target packages explicitly selected in the .config file, and
doing a two-level deep recursion in the dependencies.
With this commit, we switch legal-info to use proper make-based
dependencies, so we no longer have the limitations we used to
have. For this reason, the output of 'make legal-info' after this
patch may contain *more* entries than before this patch, but it is
really because it is now correct.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
For clarity, this commit renames the TARGETS variable to the more
meaningful PACKAGES variable. Indeed, only packages (handled by one of
the package infrastructures) should be listed in this variable, and
not other random non-package targets.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
The main Makefile was declaring a subset of the per-package targets as
being PHONY, but not all of them. Now that the pkg-generic package
infrastructure is taking care of that in a much more systematic
fashion, this commit gets rid of the unneeded code from the main
Makefile.
[Thomas: re-add list-defconfigs to the list of PHONY targets, as
noticed by Yann.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
In 8a46d4bf1f the randpackageconfig and allpackageyesconfig were
extended with disabling all the legacy options, otherwise the resulting
config couldn't be built. However, that didn't work for randconfig and
allyesconfig.
This commit reverts 8a46d4bf1f and replaces it with a different
approach: skipping of the legacy config options is passed explicitly
through the environment variable SKIP_LEGACY, which forces
BR2_SKIP_LEGACY to y. The new option BR2_SKIP_LEGACY completely
disables the legacy handling, which effectively removes all the legacy
options from the .config.
However, in that case no values are set for the legacy options so a
subsequent oldconfig will query them. Therefore, run an additional
olddefconfig.
[Peter: fix s/BR2_LEGACY/SKIP_LEGACY/ typo]
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
[yann.morin.1998@free.fr: do not rely on a user-visible option, works
perfectly well with only blind options set from the environment]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This definition of HOSTFC is completely wrong.
"$(HOSTLD)" should be "$(HOSTFC)". Also, "echo" always succeeds, so
"which g77 || type -p g77 || echo gfortran" is never run.
Anyway, HOSTFC is most likely set to "/use/bin/ld" and nobody has
complained about it before me, so I guess it is not used at all.
At least grepping HOSTFC, FC_FOR_BUILD did not hit any packages.
Drop HOSTFC and FC_FOR_BUILD.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Reviewed-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
[Thomas: fix issues noticed by Arnout:
- Rewrap the linux/Config.in paragraph
- Revert the "is a toolchain dependency" -> "has a toolchain
dependency" change from pkg-generic.mk, as the original was
correct.]
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The GNU make's origin function know undefined variable well,
so the outer ifdef/endif conditional checking is unneeded.
>From `info make` documentation, origin will return
`undefined'
if VARIABLE was never defined.
`command line'
if VARIABLE was defined on the command line.
...
Therefore, $(origin V) will get a value anyway, killing ifdef/endif
is viable and safe.
Furthermore, I've checked the minimal requirements from the top
Makefile is GNU make 3.81, and that version of GNU make has support
of origin function well already, so now it's safe to kill the outer
conditional checking, without upgrading the minimal requirements.
Signed-off-by: Cheng Renquan <crq@kernel.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
[ Commit description is borrowed from Linux Kernel
(commit b8b0618cf6fa) and adjusted for Buildroot ]
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The global .PHONY specification in the main Makefile is describing a
number of directory targets as being PHONY, which doesn't make much
sense. PHONY targets are targets that do not exist on the filesystem,
and which make should always consider as not being up-to-date, so that
the commands associated to these targets are always executed by make
when the command is invoked, even if a file with the same name exists
on the filesystem.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
This comment is a left-over from previous code changes, and it doesn't
make much sense where it is placed.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
The documentation of the package-specific targets is heavily inspired by
the earlier patch by Thomas Petazzoni [1], but the <pkg>-install*
targets were left out since they're not useful for the developer IMHO.
The *-menuconfig target's help text is moved to the package-specific
targets.
[1] http://patchwork.ozlabs.org/patch/433692
[Thomas: remove double <pkg>-graph-depends help.]
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
As discussed in the FOSDEM2015 BR developer meeting, the output of
'make help' is too long for comfortable reading. To shorten it, split
off the list of defconfigs in a new target, 'list-defconfigs'.
Declare the new target as phony.
Add 'list-defconfigs' to the documentation.
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Currently, a 'make clean' leaves the graphs/ subdirectory in the
output directory. This commit defines a GRAPHS_DIR variable, used by
the different graph-generating targets, and which gets cleaned up in
the 'clean' target.
[Thomas: use the new GRAPHS_DIR variable in more places, as suggested
by Yann.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Special characters in files or directories in the rootfs can cause
problems when stripping files. For example "target/some song.mp3"
gets treated as two entries. "target/some" and "song.mp3" are both
passed to $(STRIPCMD). This then errors saying files don't exist.
Additionally a ' and possibly other special characters in a file path
causes xargs to give the error: "xargs: unmatched single quote; by
default quotes are special to xargs unless you use the -0
option". This also has the effect of removing this entry and further
entries from the list of files to strip. This can be demonstrated by
having a test directory with the files: "cat" "rabbit's"
"elephant". then running the command: "find -name "*" -print | xargs"
To fix this we pass -print0 to find which seperates entries with a
NULL character, and we pass -0 to xargs to tell it to only use NULL
characters as the deliminator.
Signed-off-by: Andrew Parlane <andrewp@carallon.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Updated text to say that the defconfig is written to the BR2_DEFCONFIG
location.
Signed-off-by: Frank Hunleth <fhunleth@troodon-software.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
We have the BR2_DEFCONFIG feature that saves the path to the defconfig
file that was specified on the command line, so that a later
savedefconfig would immediately save to the right location. This wasn't
done for the defconfigs in the configs/ directory, however, to avoid
accidentally overwriting them.
Now we decided that it would be more useful to overwrite the defconfigs
in the configs/ directory after all. To do this, we pass the path to
that defconfig in the environment.
Cc: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Tested-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Buildroot unexports PKG_CONFIG_PATH in the top-level Makefile for purity
reasons. But it has an unfortunate side-effect in that "make menuconfig"
will not (necessarily) be able to pick up ncurses via host pkg-config,
breaking "make menuconfig" on systems where ncurses is installed in a
non-standard location.
This patch saves the original PKG_CONFIG_PATH variable in
HOST_PKG_CONFIG_PATH and restores the original PKG_CONFIG_PATH variable
only in the sub-processes that builds the various menuconfig/nconfig/...
targets.
(PKG_CONFIG_PATH has to be placed in front of the make command so that it
propagates to sub-processes. If given as an argument, it doesn't work.)
Signed-off-by: Bjørn Forsman <bjorn.forsman@gmail.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Because it's just checking the presence of the "s" character even a
make --warn-undefined-variables
is detected as a silent build.
Fix that by filtering out long options.
Signed-off-by: Fabio Porcedda <fabio.porcedda@gmail.com>
Cc: Arnout Vandecappelle <arnout@mind.be>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The comment language may lead the reader to think that .br-external is removed
whenever BR2_EXTERNAL is not set in the command line. Make it clear that
BR2_EXTERNAL must be explicitly set to an empty value for .br-external to get
removed.
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
