Extend the packageauxvars database to keep a list of possible package
dependencies for each provider, then utilize this information in buildroot
to resolve the ABI version dependencies of dependent packages up to five
levels deep.
This should properly trigger rebuilds for packages indirectly depending
on other packages whose ABI_VERSION changed.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
(cherry picked from commit 27a4a71c24)
This fixes two issues with cleaning package files from STAGING_DIR:
* CleanStaging currently can only remove files and not directories. This
changes CleanStaging to use clean-package.sh, which does remove
directories.
* Because of the way directories are ordered in the staging files list,
clean-package.sh currently tries (and fails) to remove parent
directories before removing subdirectories. This changes
clean-package.sh to process the staging files list in reverse, so that
subdirectories are removed first.
Signed-off-by: Jeffery To <jeffery.to@gmail.com>
When CONFIG_AUTOREMOVE is enabled, the build system touches an additional
".autoremove" stamp file in the cleaned build directory.
Since the autoremove stamp file is touched after the compile process
completed, it ends up being the most recent file of the package build
directory, causing the timestamp.pl check of depends.mk to erroneously
declare the ".built" and ".prepared" stamp files as stale, triggering
a forced clean-build submake process.
Fix the problem by using the ".built" stamp file as modification time
reference when touching the ".autoremove" stamp file.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Some packages, e.g. busybox, explicitly remove old .configured stamps
before attempting configuration, rather than after the actual configuration
step. This seems like a good idea, as there will be no stamp left if
configuration fails. Change generic rules to work like this, so package-
specific rules can be dropped.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
Recent attempts to use it have shown that it does not work properly
except for a few undocumented cases. It's better to remove this now to
avoid having more people fall into the same trap
Signed-off-by: Felix Fietkau <nbd@nbd.name>
quilt.mk needs to be included first, to ensure that STAMP_PREPARED does
not include the hash if quilt is used.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Instead of using a separate .clean stamp file, remove the install info
file on compile, then append the install package list afterwards
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Since the main stamp file depends on the _check stampfile, the _check
stampfile needs to be created on the first run as well.
Fixes spurious rebuilds with CONFIG_AUTOREBUILD=y
Signed-off-by: Felix Fietkau <nbd@nbd.name>
This avoids repeatedly unpacking and rebuilding packages that are
failing the build. Re-running the failing step should be much faster.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Calling the clean target removes all .ipk files and un-stages the
package. Add a new target just for clearing the build dir and call that
one instead of the full clean target
Signed-off-by: Felix Fietkau <nbd@nbd.name>
This is used to save space on buildbot instances.
If any part of a package needs to be rebuild, the whole package is
rebuilt from scratch. Stamp files are preserved to allow dependency
checks to work
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Using a single host package staging dir (and build dir) significantly speeds up
builds when multiple targets are built in succession, especially for large host
packages like NodeJS.
$(STAGING_DIR)/host is kept in addition to $(STAGING_DIR_HOSTPKG) in most
places; it is still used as destination for host files in Build/InstallDev.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
This makes it easier to unify versioning of git based package downloads.
PKG_SOURCE_DATE along with an 8-character abbreviation of the git hash
is used as PKG_VERSION, PKG_RELEASE should be used like normal packages.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
This is intended to be used for a wide array of package sanity checks.
The first check that is implemented is for the hash of downloaded files.
It checks:
- Missing hash
- Use of SHA256 instead of MD5
- dl/<file> hash not matching hash in makefile
- deprecated MD5SUM variable
The deprecated MD5SUM variable check is skipped for feeds/ until OpenWrt
is updated as well
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Since we've switched to preferring SHA256 over MD5, the old variable
name is misleading. Packages using the old name remain compatible.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
We have packages with their own parts appended to standard STAMP_CONFIGURED
(mostly with an underscore character). This will render the current
STAMP_CONFIGURED_WILD setting invalid and the build system may miss a rebuild
on config change
1. Build with config A
2. Build with config B, yet .configured_A_xx did not get cleaned
3. Return to config A, but rebuild will not happen because stamp file
of config A still exists
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
Make usually passes -j and jobserver arguments as part of MAKEFLAGS. LEDE
removes MAKEFLAGS to have better control of the build, and re-adds the
jobserver arguments with -j to the Make commandline where desired.
Make 4.2+ behave differently with these arguments passed on the commandline
than in MAKEFLAGS: -j will override the jobserver argument and the job
count will be unlimited.
Moving the flags to MAKEFLAGS will need many packages to be changed and
tested; therefore, we opt for a less invasive change for now and just
remove -j for Make 4.2+, as the jobserver argument alone is enough to
enable parallel builds for these Make versions.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
Use the new repository metadata field to choose the output directory of the
final package archives.
Non-sharable packages will be placed in the per-target package directory
while the rest will be placed in a per-repository sub directory within the
$OUTPUT_DIR/packages/$CPU_TYPE/ prefix.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
This is required to allow packages to ship libraries with a different API.
Previously the linker would favor the old shared libraries in STAGING_DIR
instead of the just compiled ones in PKG_BUILD_DIR.
Avoid that problem by clearing the package staging files right after
prepare.
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
SVN-Revision: 45759
Defining STAMP_CONFIGURED statically caused it not to reflect changes to
PKG_CONFIG_DEPENDS made after including package.mk, like the additional options
added by feeds.mk for all packages.
Furthermore, as STAMP_CONFIGURED_WILDCARD was already defined recursively, the
patsubst in its definition would never match, and in consequence, the stamps
were never removed. This caused packages not to be rebuilt when they should have
been.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
SVN-Revision: 44927
On systems with CONFIG_SITE in the environment (e.g. OpenSUSE) make will export
the CONFIG_SITE set in include/package.mk by default. This will cause host
builds to get the target site configuration, leading to all kinds of weirdness
(wrong pointer size, wrong endianess).
Fix this by explicitly unexporting CONFIG_SITE. The explicit export for the
target builds overrides the unexport, so the target builds will still correctly
get the site config.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
SVN-Revision: 44019
Introduce configuration options to build an "hardened" OpenWRT.
Options to enable Stack-Smashing Protection, FORTIFY_SOURCE and RELRO
have been introduced.
uClibc makefile now automatically detects if SSP support is necessary.
hostapd makefile has been fixed to use "^" as sed separator since
using a comma was problematic when using "-Wl,-z,now" and the like in
TARGET_CFLAGS.
Currently enabling SSP on user space depends on enabling SSP kernel
side, this is due to the fact that TARGET_CFLAGS are used to build
kernel modules (at least). Suggestions on how to avoid this are welcome.
Using "select" instead of "depends on" doesn't seem to work with choice
entries.
Tested with a lantiq (WBMR) router, GCC 4.8, uClibc and a subset of
the available packages.
Needs to be tested with GCC 4.9 and the remaining packages.
PIE not currently included.
Signed-off-by: Alessandro Di Federico <ale+owrt@clearmind.me>
SVN-Revision: 44005
The idea is still to enable it by default at some point
I've tested all ar71xx packages (except oldpackages) using CONFIG_ALL=y
Failing packages have been marked with PKG_CHECK_FORMAT_SECURITY:=0 for now
I can test more targets but i have no idea which are the most used
Signed-off-by: Etienne CHAMPETIER <champetier.etienne@gmail.com>
SVN-Revision: 42282
Remove leftover pieces from package-version-override.mk support
Add a new variable USE_SOURCE_DIR to use a custom build directory.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
SVN-Revision: 40527
This is a cryptographically signed message in MIME format.
udev Makefile.am has two different .pc files and put them
into two locations. Don't know if changing the udev Makefile
or changing the pkgconfig paths is the right way - someone
has to decide :)
udev and vala use an additional pkg-config dir
add it to the path
discovered because openobex did not find udev.pc
Signed-off-by: Dirk Neukirchen <dirkneukirchen@web.de>
SVN-Revision: 39792
