- This commit adds a config-option for JIT in libpcre2.
According to research published on
https://rust-leipzig.github.io/regex/2017/03/28/comparison-of-regex-engines/
this should give a 10x performance increase on JIT operations
which can be desireable for high performance Apache mod_rewrite
or haproxy reqrep operations.
This option is available on all officially supported architecutres
which are listed on https://pcre.org/current/doc/html/pcre2jit.html#SEC2.
Furthermore, it is enabled by default on the following
architectures: arm, i686, x86_64.
Signed-off-by: Christian Lachner <gladiac@gmail.com>
- This commit adds a config-option for JIT in libpcre(1).
According to research published on
https://rust-leipzig.github.io/regex/2017/03/28/comparison-of-regex-engines/
this should give a 10x performance increase on JIT operations
which can be desireable for high performance Apache mod_rewrite
or haproxy reqrep operations.
This option is available on all officially supported architecutres
which are listed on https://pcre.org/original/doc/html/pcrejit.html#SEC3.
Furthermore, it is enabled by default on the following
architectures: arm, i686, x86_64.
Signed-off-by: Christian Lachner <gladiac@gmail.com>
Fix breaking change introduced in the main tree with a commit
7519a36774ca ("base-files,procd: add generic service status") where the
old service `status` callback function doesn't work anymore and needs to
be renamed to `status_service`. This name was chosen for consistency
with start and stop function callbacks, which are using `start_service`
and `stop_service` naming schemes.
While at it, fix whitespace issues in the status_service as well.
Ref: http://lists.infradead.org/pipermail/openwrt-devel/2019-September/019035.html
Reported-by: Dirk Brenken <dev@brenken.org>
Signed-off-by: Petr Štetiar <ynezz@true.cz>
This way when only wanting the library nobody needs to download and
compile the server package, saving space and time. Also this way we can
avoid sudden SONAME bumps during a server upgrade.
Signed-off-by: Sebastian Kemper <sebastian_ml@gmx.net>
The issue was that the pause instruction was emitted through an asm
directive which was not correct for some mips platforms.
Simplified boost-fiber-exclude as a result.
Removed uClibc-ng math patch. It was not correct as it broke float and
long double support (std variants use function overloads). A different
solution was applied upstream. As it's quite annoying to backport, just
wait until a new release comes with that change. ARC as a platform is
barely supported anyways.
Swapped asio patch for the upstream submission, which is unfortunately, in
limbo.
Refreshed remaining patch.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
They use the same source. It makes no sense to have them separate.
Also cleaned up the Makefile to modern standards.
Fixed license information.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Cleaned up Makefile slightly. The removed CMAKE_OPTIONS are defaults from
cmake.mk
Removed Upstreamed patches.
Rebased and added .patch to the remaining one.
Added -Wformat-security patch.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
On GCC9, it throws a Wformat-nonliteral error. Unfortunately, there's no
easy was to fix it as it is fortify-headers where the warning ultimately
comes from.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
This package fails compilation as it errors on a false strncpy length
parameter under GCC9.
This package is also unmaintained as well as dead upstream. Instead of
keeping it on life support, kill it.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Several Makefile cleanups for consistency between packages.
Removed PKG_INSTALL as it's implied by CMAKE_INSTALL.
Removed InstallDev for the same reason.
Removed upstreamed patches.
Fixed license information.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
This is an alternate AF_ALG engine for openssl, based on the devcrypto
engine, but using the AF_ALG interface instead of /dev/crypto.
It is different than the AF_ALG engine that ships with OpenSSL:
- it uses sync calls, instead of async
- it suports more algorithms
Signed-off-by: Eneas U de Queiroz <cotequeiroz@gmail.com>
Added size optimizations.
Added explicit options to make sure Boost doesn't get included.
Several Makefile cleanups.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Switch to CMake. Allows to simplify the Makefile.
Replaced InstallDev section with CMAKE_INSTALL.
Added PKG_BUILD_PARALLEL for faster compilation.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
On the OpenWrt build bots the compile currently fails:
cc1: error: /builder/shared-workdir/build/sdk/staging_dir/target-arm_cortex-a5+vfpv4_musl_eabi/include: No such file or directory [-Werror=missing-include-dirs]
cc1: all warnings being treated as errors
(Or rather, it _would_ fail like this if libmbim compiled successfully.)
The directory in question gets added to TARGET_CPPFLAGS by rules.mk.
Nothing usually gets installed to that directory, so unsurprisingly it
may not exist on the build bots.
On first glance this problem cannot be reproduced. But once autoreconf
is called this becomes possible.
If called without "--enable-more-warnings", configure adds -Werror to
the flags. This commit adds "--enable-more-warnings=yes" to the build.
This way the extra warnings are kept in place, but they aren't turned
into errors no more.
"PKG_FIXUP:=autoreconf" is also added so that it's easier to forecast
build failures on the build bots when preparing future pull request.
Last but not least "--disable-silent-rules" is added, because more
output is usually welcome for debugging, be it on the build bots or at
home/the office.
Signed-off-by: Sebastian Kemper <sebastian_ml@gmx.net>
On the OpenWrt build bots the compile currently fails:
make[8]: Entering directory '/builder/shared-workdir/build/sdk/build_dir/target-arm_cortex-a5+vfpv4_musl_eabi/libmbim-1.20.0/src/common'
CC libmbim_common_la-mbim-common.lo
cc1: error: /builder/shared-workdir/build/sdk/staging_dir/target-arm_cortex-a5+vfpv4_musl_eabi/include: No such file or directory [-Werror=missing-include-dirs]
cc1: all warnings being treated as errors
The directory in question gets added to TARGET_CPPFLAGS by rules.mk.
Nothing usually gets installed to that directory, so unsurprisingly it
may not exist on the build bots.
On first glance this problem cannot be reproduced. But once autoreconf
is called this becomes possible.
If called without "--enable-more-warnings", configure adds -Werror to
the flags. This commit adds "--enable-more-warnings=yes" to the build.
This way the extra warnings are kept in place, but they aren't turned
into errors no more.
"PKG_FIXUP:=autoreconf" is also added so that it's easier to forecast
build failures on the build bots when preparing future pull request.
Last but not least "--disable-silent-rules" is added, because more
output is usually welcome for debugging, be it on the build bots or at
home/the office.
Signed-off-by: Sebastian Kemper <sebastian_ml@gmx.net>
Currently only xml2-config is installed, for both the normal libxml2
package as well as the host package. The problem with that is that due
to multilib considerations the build host may have xml2-config installed
with a host triplet prefix, like x86_64-pc-linux-gnu-xml2-config (and
xml2-config as a symbolic link to it). Gentoo for instance sets it up
like this.
Packages may actually search for a prefixed xml2-config before searching
for xml2-config. An example would be Asterisk:
checking for x86_64-pc-linux-gnu-xml2-config... /usr/bin/x86_64-pc-linux-gnu-xml2-config
This then introduces wrong information into the build, for instance
bad includes:
~/tmp/openwrt $ /usr/bin/x86_64-pc-linux-gnu-xml2-config --cflags
-I/usr/include/libxml2
When the intention is to use OpenWrt's own (host) libxml2 one would like
to see this output used instead:
~/tmp/openwrt $ ./staging_dir/hostpkg/bin/xml2-config --cflags
-I/home/sk/tmp/openwrt/staging_dir/hostpkg/include/libxml2
This commit addresses this by installing xml2-config with a suitable
prefix and creating a symbolic link xml2-config. This is done for both
the host package and the normal package. The latter also needs this fix
because the target may use the same triplet as the host system (for
instance x86_64 cross-compiling for x86_64).
Signed-off-by: Sebastian Kemper <sebastian_ml@gmx.net>
The host build requires libxml2-dev. This commit adds the dependency for
libxml2/host, as the host system may not have it installed. This also
avoids using the host's xml2-config (in /usr/bin for instance) while
linking to libraries in staging_dir/hostpkg.
ldd staging_dir/hostpkg/bin/xsltproc | grep xml2
libxml2.so.2 => /home/sk/tmp/openwrt/staging_dir/hostpkg/lib/libxml2.so.2 (0x00007fcc0644c000)
Signed-off-by: Sebastian Kemper <sebastian_ml@gmx.net>
Commit 218f0229a4 added a workaround for a
problem. The host compile overwrote xml2-config in
"$(STAGING_DIR)/host/bin". The workaround fixed the issue for regular
target packages. But it didn't leave a proper xml2-config script for
host packages.
Times have changed. Host packages are now installed in
"$(STAGING_DIR_HOSTPKG). So there is no longer any danger of the host
package overwriting the xml2-config script for target packages. So
revert the mentioned commit, leaving us with two proper xml2-config
scripts, one for target package builds and one for host package builds.
Signed-off-by: Sebastian Kemper <sebastian_ml@gmx.net>
nls.mk is needed.
explicitly disabled udev. It's not available for use.
Added autoreconf to try to fix the buildbots.
Minor cleanups.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Rosen Penev
Christian Lachner
Petr Štetiar
Sebastian Kemper
DENG Qingfang
Jiri Slachta
Yousong Zhou
Nicholas Smith
Jan Pavlinec
W. Michael Petullo
Andre Heider
Alexander Ryzhov
Eneas U de Queiroz
Daniel Golle
Jan Hák
Luiz Angelo Daros de Luca
Lucian Cristian
Florian Eckert
Jakub Jalowiczor