Up until now, we were defaulting to MIPS24K for all MIPS archs.
But, we should use the GENERIC target for MIPS{EL} and the MIPS24K target
for the 24kc CPU_TYPE.
Signed-off-by: Alexandru Ardelean <ardeleanalex@gmail.com>
The GENERIC target for OpenBLAS seems good enough for most configs, so try
to use it for MIPS64 and as default for other archs.
This was already being used for x86{_64} archs.
Signed-off-by: Alexandru Ardelean <ardeleanalex@gmail.com>
This commit adds a new patch which fixes build issue with Boost.Math.
Issue is described in upstream PR boostorg/math#684 [1]
[1]: https://github.com/boostorg/math/pull/684
Signed-off-by: Carlos Miguel Ferreira <carlosmf.pt@gmail.com>
This commit updates boost to version 1.77.0
More info about Boost 1.77.0 can be found at the usual place [1].
There are two new libraries in this version:
* Describe [2]: A C++14 reflection library, from Peter Dimov. Provides
macros for describing enumerators and struct/class members, and
primitives for querying this information.
* Lambda2 [3]: A C++14, dependency-free, single header lambda library, from
Peter Dimov. Allows simple function objects to be constructed via
expressions such as:
_1 + 5, _1 % 2 == 0, _1 > _2, or _1 == ' ' || _1 == '\t'.
[1]: https://www.boost.org/users/history/version_1_77_0.html
[2]: https://www.boost.org/libs/describe/
[3]: https://www.boost.org/libs/lambda2/
Signed-off-by: Carlos Miguel Ferreira <carlosmf.pt@gmail.com>
This also drops 01-remove-python-check.patch, which was added previously
to allow Github Actions to work.
Signed-off-by: Sebastian Kemper <sebastian_ml@gmx.net>
This was removed and then re-added, but defaulted off upstream. It's
still listed as "not recommended" but still absolutely essential for
many apps for performance.
Fixes: https://github.com/eclipse/mosquitto/issues/2284
Signed-off-by: Karl Palsson <karlp@etactica.com>
When libudev-zero is enabled, v4l-utils links against it, and the
package build fails due to a missing dependency:
Package v4l-utils is missing dependencies for the following libraries:
libudev.so.1
As disabling udev doesn't seem to be supported in v4l-utils, simply add
the dependency to fix the build.
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
Missing input validation of host names returned by Domain Name Servers
in the c-ares library can lead to output of wrong hostnames (leading to
Domain Hijacking).
The Common Vulnerabilities and Exposures (CVE) project has assigned the
name CVE-2021-3672 to this issue.
References: https://c-ares.haxx.se/adv_20210810.html
Signed-off-by: Petr Štetiar <ynezz@true.cz>
With a clean build environment and at least two parallel jobs
a compilation error may occur:
...
-- Looking for uv/version.h - found
libubox include dir: ULOOP_INCLUDE_DIRS-NOTFOUND
libubox libraries: ULOOP_LIBRARIES-NOTFOUND
...
CMake Error in lib/event-libs/uloop/CMakeLists.txt:
Found relative path while evaluating include directories of
"websockets-evlib_uloop":
"ULOOP_INCLUDE_DIRS-NOTFOUND"
...
Signed-off-by: Ronny Kotzschmar <ro.ok@me.com>
Use newly introduced procd_add_restart_mount_trigger to make sure
postmaster gets started only once PGDATA becomes available.
Relocate socket directory to /var/lib/postgresql to make it possible
to run postgresql inside a ujail.
Use signal for shutdown, so it works nicely with jail.
Allow multiple script in UCI 'config postgres-db' to be a list and
run them in order listed.
User more silent methods to check for db or role existence and make
it easy to create several databases owned by the same user by passing
the same credentials multiple times.
Remove disfunctional reload handler.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
'cp' fails with a text file busy error if it tries to overwrite an
executable file that is running. 'install' unlinks the file first, so
it will not cause the problem.
Signed-off-by: Eneas U de Queiroz <cotequeiroz@gmail.com>
368014b8 Bump version number to v1.44.0, LT revision to 34:2:20
fa16e66a nghttpx: Fix max distance in weight group/address cycle comparison
40af31da nghttpx: Set connect_blocker and live_check after shuffling addresses
9e6c0685 Fix build failure
ebad3d47 Port new ngtcp2 map implementation
d4fd0681 Bump llhttp to 6.0.2
43a47aa0 Do not return HPE_USER from where it is prohibited
20079b4c Update bash_completions
2aeec770 Update manual pages
cef458c3 Replace black-list with block-list
617a5766 Replace master with main
f1d67335 Initialize Config rps field
5f3bcb1f Add precious variables for libev and jemalloc and use JEMALLOC_CFLAGS
b419bfd9 Remove unused field
e406a2c1 Update doc
962a75c4 Merge pull request #1559 from nghttp2/h2load-rps
6cdc13d6 h2load: Add --rps option
92944f78 h2load: Allow unit in -D option
276792a8 Remove unnecessary function
579fa6ea Add more --with-* configure flags
2f2b2117 Add LIBTOOL_LDFLAGS configure variable
88a3cb51 Bump llhttp to 4.0.0
40679cf6 Merge pull request #1553 from nghttp2/nghttpx-fix-accesslog-method
5b587e85 Merge pull request #1550 from jktjkt/docs-asio-fix-typos
50a1121d nghttpx: Remove trailing white space after $method log variable
3239c5ef Bump up version number to v1.44.0-DEV
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
Remove both patches as they have been upstreamed.
This new release is only available on Github and does not ship configure
or Makefile, so autoreconf is needed.
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
uloop is always available on OpenWrt, so build in LWS's uloop support by
default. Size difference of .ipk is 683 bytes, or about 0.5%.
No known applications use this at present, but it seems better to
proactively offer it.
Signed-off-by: Karl Palsson <karlp@etactica.com>
Initial draft PR is:
https://github.com/openwrt/packages/pull/11894
This one is a bit more complete, and follows packaging practices.
For now, disabling builds on ARC and PowerPC. Will require more work to get
them going.
Explicitly disabling OpenMP support, so that it doesn't get picked by
accident.
Later we may use the `CPU_TYPE` parameter to tweak things a little further.
Signed-off-by: Alexandru Ardelean <ardeleanalex@gmail.com>
❯ alsamixer
ALSA lib conf.c:827:(get_char_skip_comments) Cannot access file /usr/share/alsa/ctl/default.conf
ALSA lib conf.c:1994:(_snd_config_load_with_include) _toplevel_:61:26:No such file or directory
ALSA lib conf.c:4040:(config_file_open) /usr/share/alsa/cards/aliases.conf may be old or corrupted: consider to remove or fix it
ALSA lib conf.c:3962:(snd_config_hooks_call) function snd_config_hook_load returned error: No such file or directory
ALSA lib control.c:1481:(snd_ctl_open_noupdate) Invalid CTL cards.ctl.default
cannot open mixer: No such file or directory
Signed-off-by: Javier Marcet <javier@marcet.info>
This backports an upstream fix for the following warning:
Warning: Unparsable number `xu \b, dcrypt version %d'
Signed-off-by: Martin Schiller <ms@dev.tdt.de>
Notable changes:
* Use toml as new default storage backend:
Upstream has removed the ini plugin, we have been using in the
past. toml is still somewhat experimental, but upstream has
designated it as future recommended default.
* Remove the dependency on boost. The only plugin needing that has
been removed upstram.
* Enable plugins for libev and libuv event loop integration. This
adds two new packages with the respective dependencies.
Upstream has fixed the bugs preventing us from using this.
* Enable the internal notification plugin/system. This allows
applications to automatically receive notifications about changes
to their configuration. Again upstream fixing bugs enables this
for us.
* Set the environment variable $XDG_CACHE_HOME to "/tmp/" globally
to prevent elektra (and possibly other applications) from
writing temporary data to flash.
Signed-off-by: Harald Geyer <harald@ccbib.org>
Python supports both GNU dbm and Berkley DBM.
GNU dbm also has a compatibility layer for Berkeley.
The current Berkley DB in OpenWrt is 4.7 and hasn't been updated in ages.
It's also pretty big.
So, an alternative for Python is to use GNU dbm for both (GNU and Berkley).
Also, removing static shared libraries from the build, to encourage the
usage of the shared ones.
Signed-off-by: Alexandru Ardelean <ardeleanalex@gmail.com>
Switch to AUTORELEASE for simplicity.
Disable parallel compilation as there's something wrong with NSS' build
system. Reliably fails with make -j 12 on a ryzen 3600.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
configure.in checks for "ARMv8 CRC32C intrinsics" and goes as far as
adding "-march=armv8-a+crc" to the target flags if the compiler allows
it. This can clash with the OpenWrt target flags in
CONFIG_TARGET_OPTIMIZATION. If for example the latter is set to
"-mcpu=cortex-a9" the following warning is issued:
cc1: warning: switch '-mcpu=cortex-a9' conflicts with '-march=armv5t' switch
This commit prevents configure.in from adding the mentioned flag. The
addition is unwanted when cross-compiling.
An issue was raised for this recently, see [1].
[1] https://github.com/openwrt/packages/issues/16034
Signed-off-by: Sebastian Kemper <sebastian_ml@gmx.net>
I'm reverting this commit, because it is pulling newer source code, but
without doing a new version bump. This results that users who already
installed this library on their router will have different package
content than users who installed it after the commit.
The another small reason is that commit subject - update revision is cryptic itself and should be more bulletproof like saying update to version 1.1
More details: https://github.com/openwrt/packages/pull/15990#issuecomment-870333695
This reverts commit e779323481.
Signed-off-by: Josef Schlehofer <pepe.schlehofer@gmail.com>
We should use upstream patch here, which can be removed later,
instead of maintaining our own one.
Reported-by: Rosen Penev <rosenp@gmail.com>
Signed-off-by: Michael Heimpold <mhei@heimpold.de>
This patch fixes the compilation issue of the upstream Cmake file.
If this patch is not applied we get the following compilation error output.
CMake Error at cmake/ConnectorName.cmake:30 (ENDMACRO):
Flow control statements are not properly nested.
Call Stack (most recent call first):
CMakeLists.txt:423 (INCLUDE)
The blamed cmake/ConnectorName.cmake file gets fixed with this patch.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
ICU 69 updates to CLDR 39 locale data with many additions and corrections. ICU 69 also includes significant improvements for measurement unit formatting and number formatting in general, as well as many other bug fixes and enhancements.
Signed-off-by: Hirokazu MORIKAWA <morikw2@gmail.com>
domoticz wants them. No idea how to avoid it. Easier to just add back
here.
Removed several outdated options from CMake times.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
In v1.76.0 Boost.Regex became a header-only library.
With this update, there are now two different versions:
- v4 for C++03 (deprecated)
- v5 header-only
This commit fixes an issue which was preventing Boost.Regex
from being built for old ArmV5 targets
Signed-off-by: Carlos Miguel Ferreira <carlosmf.pt@gmail.com>
Remove myself as maintainer from PowerDNS Related packages and add
Peter van Dijk from PowerDNS as the new maintainer
Signed-off-by: James Taylor <james@jtaylor.id.au>
Remove celt051 option as upstream removed it.
Remove upstreamed patches.
Switch to AUTORELEASE for simplicity.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Building the xr_usb_serial module fails for Kenel 5.10 with CONFIG_PM
enabled:
xr_usb_serial_common.c:1574:15: error: 'ASYNCB_INITIALIZED' undeclared
(first use in this function); did you mean 'RCU_INITIALIZER'?
Use tty_port_initialized in order to determine the status of the TTY
port. This is compatible with Kernel 5.10 and at least Kernel 5.4.
Signed-off-by: David Bauer <mail@david-bauer.net>
After the recent libgpg-error update, the workaround for gpgrt-config
is no more necessary.
Reverts 63481e619 ("libassuan: fix linking of host's libgpg-error").
Signed-off-by: Alexander Egorenkov <egorenar-dev@posteo.net>
Paul Blazejowski discovered that the recent update to this package broke
cross-compiling for some OpenWrt platforms. This backports commit 33593864
(March 26, 2021) from the upstream libgpg-error project to fix this
problem.
Signed-off-by: W. Michael Petullo <mike@flyn.org>
gpgrt-config is trying to use the host's /usr/lib path when looking for
libgpg-error. Therefore, disable it and gpg-error-config will be used
as a fallback.
Signed-off-by: Alexander Egorenkov <egorenar-dev@posteo.net>
Library for node.js will be disabled temporarily due to difficulty in supporting the latest node.js.
Signed-off-by: Hirokazu MORIKAWA <morikw2@gmail.com>
Library for node.js will be disabled temporarily due to difficulty in supporting the latest node.js.
Signed-off-by: Hirokazu MORIKAWA <morikw2@gmail.com>
h2o is the library dnsdist uses to offer DNS over HTTPS to clients. dnsdist is the only user of h2o in this tree.
While h2o can depend on Ruby (to build mruby support), this is disabled in the OpenWRT build of h2o. Hence, the Ruby dependency is unnecessary, and removing it saves a few megabytes of disk space.
Signed-off-by: Peter van Dijk <peter.van.dijk@powerdns.com>
This commit updates boost to version 1.76.0
There are no new libraries in this version
More info about Boost 1.76.0 can be found at the usual place [1].
Note: This package update includes a fix merged to Boost.Fiber in [2]
which did not make into this version but it will be present in the next
one. For now, the patch is needed, but it will be removed in version
1.77.0
[1]: https://www.boost.org/users/history/version_1_76_0.html
[2]: https://github.com/boostorg/fiber/pull/276
Signed-off-by: Carlos Miguel Ferreira <carlosmf.pt@gmail.com>
This enables tools again which has been disabled in commit
498506af9 ("libupnp: update to 1.14.5"). Disabling tools leads to the
header upnptools.h not being installed into /usr/include. But e.g. mpd-full
depends on this header.
Fixes: 498506af9 ("libupnp: update to 1.14.5")
Signed-off-by: Alexander Egorenkov <egorenar-dev@posteo.net>
Removed two now pointless patches as they were added as options.
Switch to AUTORELEASE for simplicity.
Update MESON_ARGS.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
This reverts commit b29e609701.
Adding DEPENDS+=@!arc will cause a circular dependency, because some
packages select libseccomp based on a build option.
Commit e29483d7e ("libseccomp: workaround a recursive dependency") added
a workaround that was not properly documented, so I'll explain here.
The problem arises when libseccomp is selected depending on some config
option:
define Pakcage/foo
DEPENDS=+FOO_SECCOMP:libseccomp
Even if the condition is correctly defined, excluding arc, such as:
define Package/foo/config
config FOO_SECCOMP
depends on !arc
the config generator will parse libseccomp's DEPENDS variable and
generate menuconfig statements like these:
config PACKAGE_foo
select PACKAGE_libseccomp if FOO_SECCOMP
depends on !FOO_SECCOMP || !arc
The last condition is always true because FOO_SECCOMP will always be
be false when arc is true. The config generator is not able to
simplify/optimize the condition.
The circular dependecy occurs because FOO_SECCOMP depends on
PACKAGE_foo, and the redundant, always true line will make PACKAGE_foo
depend on FOO_SECCOMP.
As a workaround, we can add the 'depends on !arc' line to
Package/libseccomp/config, outside of the DEPENDS variable, so that the
redundant depends line line does not get generated.
Signed-off-by: Eneas U de Queiroz <cotequeiroz@gmail.com>
Cc: Daniel Golle <daniel@makrotopia.org>
Libdvbcsa is a free implementation of the DVB Common Scrambling Algorithm DVB/CSA - with encryption and decryption capabilities.
OpenWrt packages like `tvheadend` and `minisatip` can benefit from it.
Signed-off-by: Rafał Dzięgiel <rafostar.github@gmail.com>
faster to compile.
A small selection of packages was tested going from:
Executed in 696.30 secs fish external
usr time 82.98 mins 395.00 micros 82.98 mins
sys time 9.02 mins 0.00 micros 9.02 mins
to:
Executed in 592.20 secs fish external
usr time 84.84 mins 361.00 micros 84.84 mins
sys time 8.85 mins 57.00 micros 8.85 mins
Tested by running make -j 12 and wiping staging/build_dir/target_x
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Switch to AUTORELEASE for simplicity.
Switch to building with Ninja for faster compilation.
Refreshed patches.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Switch to AUTORELEASE for simplicity.
Switch to building with Ninja for faster compilation.
Refresh and simplify patch.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
It turns out that this is also broken on mips64el. Further testing
reveals that
-mips32r2 -mtune=mips32r2 -mabi=32
compiles the PAUSE instruction just fine whereas
-mips64r2 -mtune=mips64r2 -mabi=64
does not. The PAUSE instruction was introduced in version 2.6 of the
MIPS ISA and GCC for some reason does not allow usage of it with MIPS64.
Modify the macro to fix the situation instead of just matching on
octeon, which is not quite correct.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Rosen Penev