Upgrade mariadb from 5.5.46 to 5.5.50.
* update runtime dependency of mariadb-setupdb which requires command install
[AK]
This update appears to be mostly bug fixes and is why is was merge from master.
plus the following securiy fixes
5.5.49:
CVE-2016-0647 CVE-2016-0648 CVE-2016-0666 CVE-2016-0643
CVE-2016-5444 CVE-2016-3452
5.5.48:
CVE-2016-0640 CVE-2016-0644 CVE-2016-0646 CVE-2016-0649
CVE-2016-0650 CVE-2016-0641
5.5.47:
CVE-2016-0546 CVE-2016-0505 CVE-2016-0596 CVE-2016-0597
CVE-2016-0616 CVE-2016-0598 CVE-2016-0600 CVE-2016-0606
CVE-2016-0608 CVE-2016-0609 CVE-2016-0642 CVE-2016-0651
CVE-2016-2047
Signed-off-by: Kai Kang <kai.kang@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
(cherry picked from commit 6255219b93)
Signed-off-by: Armin Kuster <akuster808@gmail.com>
error: file /usr/share/doc/README from install of mariadb-leftovers-5.5.46-*
conflicts with file from package dmidecode-doc-2.12-*
Signed-off-by: Li Xin <lixin.fnst@cn.fujitsu.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Armin Kuster <akuster808@gmail.com>
While SYSTEMD_AUTO_ENABLE_mariadb-server = "enable", the mysqld service
hungs.
...
[ **] A start job is running for Run pending postinsts (25s / no limit)
[ OK ] Stopped MariaDB database server.
...
In mariadb-server's pkg_postinst, it install db at first runtime. And the
following 'systemctl mysqld restart' casued the hunging. So the fix idea
is to reomove pkg_postinst and still install db at first runtime.
Introduce mysql-systemd-start from ${S}/packaging/rpm-oel/. For review
convenience, we add them as file.
The mysql-systemd-start provides two functions: the install_db is to install
db at fist runtime (the first runtime means if a db existed, the install_db
will directly exit); the pinger is to wait for mysqld service startup
completed.
The mysqld.service add ExecStartPost than previous which invoke
'mysql-systemd-start post' to wait for mysqld service startup completed.
We add a package to provide install_db, so the user could choose it
to install database for mariadb at first boot before mysqld started.
It also fix another issue:
When you manually restart mysqld and do mysql test to connect the server,
the return of the restart could make sure mysqld is ready, and the following db
connect will not fail with:
...
Can't connect to local MySQL server through socket
...
Tweak my.cnf to remove obsolete/incorrect parameter.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
The libmysqlclient_r library is now a symbolic link for the
libmysqlclient one, however some code still requires it for
linking.
To easy the transition, when installing the libmysqlclient-dev we
should include the libmysqlclient-r-dev as well.
Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
This is the latest release in the 5.5.x stable series. The CVE patch has been applied upstream.
Signed-off-by: Koen Kooi <koen.kooi@linaro.org>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
When 'valgrind' is not in PACKAGECONFIG, 'WITHOUT_VALGRIND' is passed to
cmake but it is not a valid cmake macro. Warning messages show in
log.do_configure:
CMake Warning:
Manually-specified variables were not used by the project:
WITHOUT_VALGRIND
Use -DWITH_VALGRIND=FALSE instead.
Signed-off-by: Kai Kang <kai.kang@windriver.com>
Signed-off-by: Randy MacLeod <Randy.MacLeod@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Mariadb depends on openssl, openssl conditionally depends on krb5 by
checking if OPENSSL_NO_KRB5 is defined, but cmake can not handle "ifndef"
and always ask krb5.h if krb5.h exist when cmake make_depend;
Faking a kssl.h if krb5 is not in PACKAGECONF, which does not include
krb5 related declaration, install the faked kssl.h into build dir to
be searched firstly
Signed-off-by: Roy Li <rongqing.li@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Support i686 without needing to duplicate the i586 over-ride.
Signed-off-by: Andre McCurdy <armccurdy@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
1. change the datadir from /var/mysql to /var/lib/mysql
2. after creating datedir, restorecon on it, make SELinux be able to
work
3. Add the PIDFile for systemd unit file, otherwise systemctl can not
stop mysqld
Signed-off-by: Roy Li <rongqing.li@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
/etc/init.d/mysqld from mariadb-server uses mysqladmin, which is installed
into mariadb-client, so add the dependency to make this script be able
to run
Signed-off-by: Roy Li <rongqing.li@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
mariadb failed to build:
| make[2]: *** No rule to make target
'/.../bitbake_build/tmp/sysroots/qemux86-64/usr/lib64/libevent.so',
needed by 'tests/async_queries'. Stop.
| make[2]: Leaving directory
'/.../bitbake_build/tmp/work/core2-64-wrs-linux/mariadb/5.5.40-r0/build'
CMakeFiles/Makefile2:6742: recipe for target 'tests/CMakeFiles/async_queries.dir/all' failed
| make[1]: *** [tests/CMakeFiles/async_queries.dir/all] Error 2
It is an rare case that when configure mariadb, libevent exists in
sysroot. But when do compile for mariadb, libevent is gone somehow. That
causes build fails. Run follow commands could reproduce this error:
$ bitbake libevent
$ bitbake mariadb -c cleansstate
$ bitbake mariadb -c configure
$ bitbake libevent -c clean
$ bitbake mariadb
Add if without libevent, test case tests/async_queries.c will not be built.
The case is helpful to do unit test. So add libevent as a dependency to fix the
error and make sure async_queries.c will be built.
Signed-off-by: Kai Kang <kai.kang@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
There are 3 configure files under /etc/my.cnf.d:
client.cnf: options for client library
mysql-clients.cnf: options for MariaDB tools such as mysqladmin, etc
server.cnf: options for server
They are packaged to mariadb-leftover at this moment. That is not correct.
Split them into sub-packages:
client.cnf --> libmysqlclient
mysql-clients.cnf --> mariadb-client
server.cnf --> mariadb-server
Make these files as configure file by CONFFILES and update global
configure file my.cnf to include /etc/my.cnf.d as well.
Signed-off-by: Kai Kang <kai.kang@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
* rename mariadb_5.5.39.inc -> mariadb.inc
* add dependency bision-native for mariadb-native
* add revert-fix-for-MDEV-5120.patch to revert an commit for mariadb test suite
which causes packages mysql-python and modphp fail to build
* add PACKAGECONFIG 'libedit'
Signed-off-by: Kai Kang <kai.kang@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Changes:
- Add patch to check valgrind headers only if WITH_VALGRIND is set
- Add PACKAGECONFIG for valgrind and disable it by default
Signed-off-by: Jackie Huang <jackie.huang@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Use ${PN} instead of hardcoding 'mariadb' so that everything works
fine in case of multilib.
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Change socket location to /var/lib/mysql/mysql.sock. Otherwise, we
could spot errors in systemd systems like below.
mysqladmin: connect to server at 'localhost' failed
error: 'Can't connect to local MySQL server through socket '/tmp/mysql.sock' (2)'
Check that mysqld is running and that the socket: '/tmp/mysql.sock' exists!
This is because mysqld.service is using PrivateTmp=true.
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
The anonymous function redefined bindir to bindir_noprefix (and
others), which broke the sstate between build dirs.
The redefined *_noprefix was to used to pass to the INSTALL_*
variables to change the layout as we want, but in fact we can
do this in other ways instead of the pain one:
1) Change the default layout and a little adjustment.
There are 4 install layout: STANDALONE(default), RPM, DEB and SVR4.
And RPM is the one close to what we are using.
2) Don't use prefix and pass full paths to INSTALL_*.
The mariadb's cmake define some of the INSTALL_* relative to
CMAKE_INSTALL_PREFIX, So we can use empty CMAKE_INSTALL_PREFIX,
then we can pass our full paths to INSTALL_* directly.
This patch set the default layout to RPM and pass paths only for:
- INSTALL_DOCDIR: no prefix prepending, so ${datadir} works.
- INSTALL_LIBDIR/INSTALL_PLUGINDIR: use ${baselib}.
The mysql-test is moved from /usr to /usr/share which is more
reasonable, fix the FILES inclusion accordingly.
Signed-off-by: Jackie Huang <jackie.huang@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Upgrade mariadb from 5.5.38 to 5.5.39. Previous upgrade made a lot of
work, so it is easy this time.
Signed-off-by: Kai Kang <kai.kang@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
The service file mainly comes from Fedora20.
This patch also adds a configuration file under /etc/tmpfiles.d so
that mysqld could start correctly.
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
In ubuntu/redhat, we have:
...
ps -eo user,group,cmd | grep "mysql"
mysql mysql /usr/sbin/mysqld
...
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
* Upstream switched from autotools to cmake
* Separate build dir and parallel make now work
* Library versioning no longer used for plugins; other libtool cruft
gone
* Proper upstream initscript
* Plugins moved from client library package to mariadb-server package
(matches how Fedora packages these)
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Perl and perl modules don't be installed in small image, but mariadb-server, mariadb-client
and mariadb-leftovers need them in runtime, so we add them to RDEPENDS.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Chong Lu <Chong.Lu@windriver.com>
Changes:
- rename SUMMARY with length > 80 to DESCRIPTION
- rename DESCRIPTION with length < 80 to (non present tag) SUMMARY
- drop final point character at the end of SUMMARY string
- remove trailing whitespace of SUMMARY line
Note: don't bump PR
Signed-off-by: Matthieu Crapet <Matthieu.Crapet@ingenico.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
When using a recent cross-binutils you need this fix as well, so move it from -native to the .inc.
Signed-off-by: Koen Kooi <koen.kooi@linaro.org>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Add patch to tell autoconf about with-plugin option.
In this way we avoid warnings like:
configure: WARNING: unrecognized options: --with-plugin-maria
Change-Id: Id64ac01d7f2072e1bca8979c5da3e8c6520105cb
Signed-off-by: Andrei Gherzan <andrei@gherzan.ro>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
* unlike BBCLASSEXTENDed native support, dependencies in mariadb-native don't get
automatic -native suffix, so mariadb-native was depending on target ncurses and zlib.
* move the dependency from .inc and apply it with right suffix
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Below errors only occurs on Ubuntu 13.10:
$arch-linux-libtool: link: g++ ... -o .libs/mysqltest_embedded \
../../libmysqld/.libs/libmysqld.so -ldl
../../libmysqld/.libs/libmysqld.so: undefined reference to `dlopen'
../../libmysqld/.libs/libmysqld.so: undefined reference to `dlclose'
../../libmysqld/.libs/libmysqld.so: undefined reference to `dlerror'
../../libmysqld/.libs/libmysqld.so: undefined reference to `dlsym'
GCC/ld verion on the host:
gcc (Ubuntu/Linaro 4.8.1-10ubuntu8) 4.8.1
GNU ld (GNU Binutils for Ubuntu) 2.23.52.20130913
This is a strange behavior on Ub13.10, it fails even '-ldl' in the
link command line. This patch will append '-ldl' to dependency_libs
in libmysqld.la.
Signed-off-by: Wenzong Fan <wenzong.fan@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
bitbake lib32-mariadb
...
WARNING: QA Issue: ELF binary 'bitbake_build/tmp/work/x86-wrsmllib32-linux/lib32-mariadb/5.1.67-r0/packages-split/lib32-libmysqlclient/usr/lib/plugin/ha_xtradb.so.0.0.0' has relocations in .text
WARNING: QA Issue: ELF binary 'bitbake_build/tmp/work/x86-wrsmllib32-linux/lib32-mariadb/5.1.67-r0/packages-split/lib32-libmysqlclient/usr/lib/plugin/ha_innodb_plugin.so.0.0.0' has relocations in .text
...
The mariadb used absolute addresses on IA-32, we should forbid this.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
* Remove all PR = "r0" from all .bb files in meta-oe repo. This was done
with the command sed -e '/^PR.*=.*r0\"/d' meta*/recipes*/*/*.bb -i
* We've switching to the PR server, PR bumps are no longer needed and
this saves people either accidentally bumping them or forgetting to
remove the lines (r0 is the default anyway).
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Mariadb will provide zlib.* if it can't find zlib binaries and headers
from sysroots path. We have set it to use external zlib.* firstly with
these definitions/configs in inc file:
+ DEPENDS = "ncurses zlib"
+ --with-zlib-dir=${STAGING_EXECPREFIXDIR}
But native bb file overrides the dependencies with:
+ DEPENDS = "ncurses-native"
If mariadb-naitve was built before zlib-native, its own zlib will be
built and provide zlib.la which will brake other packages link to real
zlib with errors like:
* .../usr/lib/libz.a(gzclose.o): relocation R_X86_64_32 against \
`a local symbol' can not be used when making a shared object; \
recompile with -fPIC
* .../usr/lib/libz.a: could not read symbols: Bad value
By removing the DEPENDS = "ncurses-native" from the native bb file,
mariadb-native will still have the DEPENDS (ncurses zlib) from the
inc file, and since the recipe is mariadb-native, this is equivalent
to DEPENDS = "ncurses-native zlib-native" due to the map_dependencies()
inner fuction for native_virtclass_handler() in "oe-core/meta/classes/
native.bbclass".
Signed-off-by: Wenzong Fan <wenzong.fan@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
It failed to build image while IMAGE_INSTALL += "mysql5"
For backwards compatibility:
- Add mysql5 to RPROVIDES_${PN}
- Add mysql5-dbg to RPROVIDES_${PN}-dbg
- Add mysql5-leftovers to RPROVIDES_${PN}-leftovers
- Use "+=" instead of "=" in RPROVIDES_${PN}-client and
RPROVIDES_${PN}-server
The same to RREPLACES and RCONFLICTS.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Here is the error log from mariadb's configure test
...
checking for zlib compression library... configure: error: headers or
binaries were not found in /home/jiahongxu/yocto/build-20131009-mysql/
bitbake_build/tmp/sysroots/qemux86-64/usr/{include,lib}
...
When multilib is used, the baselib is assigned with "lib64", here is
the log from 'bitbake mariadb -e'
...
4738 # $baselib [3 operations]
4739 # set conf/bitbake.conf:10
4740 # "${BASELIB}"
4741 # set conf/bitbake.conf:11
4742 # [vardepvalue] "${baselib}"
4743 # set /home/jiahongxu/yocto/build-20131009-mysql/layers/oe-core/meta/conf/multilib.conf:2
4744 # "${@d.getVar('BASE_LIB_tune-' + (d.getVar('DEFAULTTUNE', True) or 'INVALID'), True) or d.getVar('BASELIB', True)}"
4745 # computed:
4746 # "${@d.getVar('BASE_LIB_tune-' + (d.getVar('DEFAULTTUNE', True) or 'INVALID'), True) or d.getVar('BASELIB', True)}"
4747 baselib="lib64"
...
In this situation, the zlib locates in 'lib64' rather than 'lib'.
But mariadb's confiure test still searches zlib in 'lib'.
Modify mariadb's confiure to let lib dir configurable rather
than hardcode could fix this issue.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Invoke 'mysql_install_db' on target, there are some warnings:
...
Installing MariaDB/MySQL system tables...
131009 6:13:14 [Warning] '--skip-locking' is deprecated and will be removed in a future release. Please use '--skip-external-locking' instead.
OK
Filling help tables...
131009 6:13:14 [Warning] '--skip-locking' is deprecated and will be removed in a future release. Please use '--skip-external-locking' instead.
OK
...
Modify my.cnf to use '--skip-external-locking' instead of '--skip-locking'
could fix this issue.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
In mariadb recipe,do_install move all the lib files from usr/lib/mysql
to usr/lib,but didn't change libdir value in *.la libtool file. This
will cause libtool can't get correct lib path of mysql when other
package need mysql lib.So update them to correct value.
Here is a exapmle:
In libmysqlclient.la file: libdir = "/usr/lib/mysq/", but in fact
libmysqclient.so is installed into "/usr/lib", this will cause
libtool can not find correct lib file with libmysqlclient.la, then
libtool will use rpath option to reference a absolute path, and
this will cause rpath WARING and rpm error.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Switch to MariaDB (which is a drop-in replacement for MySQL) and use
the latest stable release from the 5.1 series.
* Update LIC_FILES_CHKSUM due to reformatted GPLv2 license text with
updated FSF address
* Refresh patches
* Add two patches from the upstream 5.1 branch to fix CVE-2013-1861
* Add a package for libmysqld (the embedded server library)
* Disable "maria" plugin since this fails to compile and doesn't
appear to be critical
* Drop some unrecognised/redundant options from EXTRA_OECONF
* Fix text relocation QA warnings introduced in the upgrade
* Convert to use useradd.bbclass for creating mysql user
* Set SUMMARY instead of description
* Move SRC_URI checksums to the version-specific inc file
* Clear out cruft in files/
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
* fails without mysql_config:
ERROR: Function failed: do_compile (log file is located at mysql-python/1.2.4-r0/temp/log.do_compile.29833)
ERROR: Logfile of failure stored in: mysql-python/1.2.4-r0/temp/log.do_compile.29833
Log data follows:
| DEBUG: Executing shell function do_compile
| sh: 1: mysql_config: not found
| Traceback (most recent call last):
| File "setup.py", line 18, in <module>
| metadata, options = get_config()
| File "mysql-python/1.2.4-r0/MySQL-python-1.2.4/setup_posix.py", line 43, in get_config
| libs = mysql_config("libs_r")
| File "mysql-python/1.2.4-r0/MySQL-python-1.2.4/setup_posix.py", line 25, in mysql_config
| raise EnvironmentError("%s not found" % (mysql_config.path,))
| EnvironmentError: mysql_config not found
| ERROR: python setup.py build_ext execution failed.
| ERROR: Function failed: do_compile (log file is located at mysql-python/1.2.4-r0/temp/log.do_compile.29833)
NOTE: recipe mysql-python-1.2.4-r0: task do_compile: Failed
ERROR: Task 6 (meta-oe/recipes-support/mysql/mysql-python_1.2.4.bb, do_compile) failed with exit code '1'
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
* in some seldom cases mysql's m4 macro decides that system
zlib (from DEPENDS) is not worthy and decides to use own
bundled version which prefers stuff like static libz.a and
stages libz.la again:
mysql-5.1.40/config/ac-macros/zlib.m4
NOTE: recipe mysql5-5.1.40-r11: task do_package_setscene: Started
NOTE: recipe zlib-1.2.7-r0: task do_packagedata_setscene: Succeeded
WARNING: The recipe zlib is trying to install files into a shared area when those files already exist. Those files and their manifest location are:
sysroots/om-gta04/usr/lib/libz.a
Matched in manifest-om-gta04-mysql5
Please verify which package should provide the above files.
NOTE: recipe zlib-1.2.7-r0: task do_populate_sysroot_setscene: Succeeded
* this libz.la + libz.a is breaking many packages which are now trying to
link shared libs against libz.a from mysql which was built without fPIC
breaking libxml2, glib-2.0, gnutls, libpng, ...
| ld: sysroots/om-gta04/usr/lib/libz.a(crc32.o): relocation R_ARM_MOVW_ABS_NC against `a local symbol' can not be used when making a shared object; recompile with -fPIC
| sysroots/om-gta04/usr/lib/libz.a: could not read symbols: Bad value
| collect2: ld returned 1 exit status
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
When mysql does not detect zlib during configure then their copy of
zlib 1.2.3 is built. That leads to problems during rootfs creation:
| * check_data_file_clashes: Package libz-staticdev wants to install file /home/hrw/HDD/devel/canonical/aarch64/openembedded/build/tmp-eglibc/work/genericarmv8-oe-linux/linaro-image-lamp/1.0-r0/rootfs/usr/lib/libz.a
| But that file is already provided by package * libmysqlclient-staticdev
| * opkg_install_cmd: Cannot install package libz-staticdev.
Signed-off-by: Marcin Juszkiewicz <marcin.juszkiewicz@linaro.org>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Kai Kang
Li xin
Koen Kooi
Hongxu Jia
Randy MacLeod
Otavio Salvador
Roy Li
Koen Kooi
Mario Domenech Goulart
Andre McCurdy
Paul Eggleton
Jackie Huang
Chen Qi
Chong Lu
Matthieu CRAPET
Andrei Gherzan
Martin Jansa
Wenzong Fan
Marcin Juszkiewicz