From dc34413ab6bbf32bcaeaca6437534285981909c2 Mon Sep 17 00:00:00 2001 From: Remi Collet Date: Tue, 14 Feb 2012 19:26:38 +0100 Subject: MySQL : sync with latest rawhide changes --- mysql-disable-test.patch | 27 ---------------- mysql-install-test.patch | 41 +++++++++++++++++------- mysql-logrotate.patch | 25 +++++++++++++++ mysql55.spec | 80 ++++++++++++++++++++++++++++++++++------------ mysqld.service | 16 ++++++++++ rh-skipped-tests-arm.list | 8 +++++ rh-skipped-tests-base.list | 20 ++++++++++++ 7 files changed, 159 insertions(+), 58 deletions(-) delete mode 100644 mysql-disable-test.patch create mode 100644 mysql-logrotate.patch create mode 100644 rh-skipped-tests-arm.list create mode 100644 rh-skipped-tests-base.list diff --git a/mysql-disable-test.patch b/mysql-disable-test.patch deleted file mode 100644 index 74b87df..0000000 --- a/mysql-disable-test.patch +++ /dev/null @@ -1,27 +0,0 @@ -Disable the outfile_loaddata test, which as of 5.1.38 is giving -platform-dependent results, with the "expected" results being arguably the -wrong ones. This is upstream at http://bugs.mysql.com/bug.php?id=46895 -(note that upstream has also disabled it, but only for Solaris, so we still -need this patch). -Still broken in 5.5.14, despite alleged fix. - -Also disable sys_vars.plugin_dir_basic, which is broken because -mysql-test-run doesn't set the right value of MYSQL_LIBDIR. Upstream at -http://bugs.mysql.com/bug.php?id=52223 -Still fails in 5.5.16, even though upstream marked it closed. - -Also disable innodb.innodb, which is showing platform-dependent results -as of 5.5.9. Upstream at http://bugs.mysql.com/bug.php?id=60155 - - -diff -Naur mysql-5.5.16.orig/mysql-test/t/disabled.def mysql-5.5.16/mysql-test/t/disabled.def ---- mysql-5.5.16.orig/mysql-test/t/disabled.def 2011-09-09 11:56:40.000000000 -0400 -+++ mysql-5.5.16/mysql-test/t/disabled.def 2011-10-16 19:50:10.064578284 -0400 -@@ -17,3 +17,7 @@ - archive-big : Bug#11817185 2011-03-10 Anitha Disabled since this leads to timeout on Solaris Sparc - log_tables-big : Bug#11756699 2010-11-15 mattiasj report already exists - mysql_embedded : Bug#12561297 2011-05-14 Anitha Dependent on PB2 changes - eventum#41836 -+# -+outfile_loaddata : bug#46895 code wrong, expected results wrong too -+sys_vars.plugin_dir_basic : bug#52223 fails for lib64 library directory -+innodb.innodb : bug#60155 has platform-dependent results diff --git a/mysql-install-test.patch b/mysql-install-test.patch index 5980aea..a31a280 100644 --- a/mysql-install-test.patch +++ b/mysql-install-test.patch @@ -1,27 +1,46 @@ Improve the documentation that will be installed in the mysql-test RPM. -diff -Naur mysql-5.1.43.orig/mysql-test/README mysql-5.1.43/mysql-test/README ---- mysql-5.1.43.orig/mysql-test/README 2010-01-15 12:14:43.000000000 -0500 -+++ mysql-5.1.43/mysql-test/README 2010-02-13 21:18:06.000000000 -0500 -@@ -6,6 +6,16 @@ - actually have a co-existing MySQL installation. The tests will not - conflict with it. +diff -Naur mysql-5.5.20.orig/mysql-test/README mysql-5.5.20/mysql-test/README +--- mysql-5.5.20.orig/mysql-test/README 2011-12-16 14:52:05.000000000 -0500 ++++ mysql-5.5.20/mysql-test/README 2012-02-10 17:06:19.531082253 -0500 +@@ -1,14 +1,26 @@ + This directory contains a test suite for the MySQL daemon. To run +-the currently existing test cases, simply execute ./mysql-test-run in +-this directory. It will fire up the newly built mysqld and test it. ++the currently existing test cases, execute ./mysql-test-run in ++this directory. +-Note that you do not have to have to do "make install", and you could +-actually have a co-existing MySQL installation. The tests will not +-conflict with it. +- +-All tests must pass. If one or more of them fail on your system, please +-read the following manual section for instructions on how to report the +-problem: +For use in Red Hat distributions, you should run the script as user mysql, +so the best bet is something like + cd /usr/share/mysql-test -+ sudo -u mysql ./mysql-test-run ++ sudo -u mysql ./mysql-test-run --skip-test-list=rh-skipped-tests.list +This will use the installed mysql executables, but will run a private copy +of the server process (using data files within /usr/share/mysql-test), +so you need not start the mysqld service beforehand. ++ ++The "--skip-test-list=rh-skipped-tests.list" option excludes tests that are ++known to fail on one or more Red-Hat-supported platforms. You can omit it ++if you want to check whether such failures occur for you. Documentation ++about the reasons for omitting such tests can be found in the file ++rh-skipped-tests.list. ++ +To clean up afterwards, remove the created "var" subdirectory, eg + sudo -u mysql rm -rf /usr/share/mysql-test/var + - All tests must pass. If one or more of them fail on your system, please - read the following manual section for instructions on how to report the - problem: -@@ -25,7 +35,8 @@ ++If one or more tests fail on your system, please read the following manual ++section for instructions on how to report the problem: + + http://dev.mysql.com/doc/mysql/en/mysql-test-suite.html + +@@ -25,7 +37,8 @@ With no test cases named on the command line, mysql-test-run falls back to the normal "non-extern" behavior. The reason for this is that some diff --git a/mysql-logrotate.patch b/mysql-logrotate.patch new file mode 100644 index 0000000..dd8742a --- /dev/null +++ b/mysql-logrotate.patch @@ -0,0 +1,25 @@ +Adjust the mysql-log-rotate script to contain the correct log file +name for Red Hat installations. + + +diff -Naur mysql-5.5.20.orig/support-files/mysql-log-rotate.sh mysql-5.5.20/support-files/mysql-log-rotate.sh +--- mysql-5.5.20.orig/support-files/mysql-log-rotate.sh 2011-12-16 14:52:06.000000000 -0500 ++++ mysql-5.5.20/support-files/mysql-log-rotate.sh 2012-01-26 23:24:54.402676177 -0500 +@@ -3,7 +3,7 @@ + # in the [safe_mysqld] section as follows: + # + # [safe_mysqld] +-# err-log=@localstatedir@/mysqld.log ++# err-log=/var/log/mysqld.log + # + # If the root user has a password you have to create a + # /root/.my.cnf configuration file with the following +@@ -18,7 +18,7 @@ + # ATTENTION: This /root/.my.cnf should be readable ONLY + # for root ! + +-@localstatedir@/mysqld.log { ++/var/log/mysqld.log { + # create 600 mysql mysql + notifempty + daily diff --git a/mysql55.spec b/mysql55.spec index a44dd4e..c50e773 100644 --- a/mysql55.spec +++ b/mysql55.spec @@ -1,6 +1,6 @@ Name: mysql Version: 5.5.20 -Release: 1%{?dist} +Release: 2%{?dist} Summary: MySQL client programs and shared libraries Group: Applications/Databases @@ -35,6 +35,8 @@ Source10: mysql.tmpfiles.d Source11: mysqld.service Source12: mysqld-prepare-db-dir Source13: mysqld-wait-ready +Source14: rh-skipped-tests-base.list +Source15: rh-skipped-tests-arm.list # Working around perl dependency checking bug in rpm FTTB. Remove later. Source999: filter-requires-mysql.sh @@ -47,13 +49,13 @@ Patch5: mysql-stack-guard.patch Patch6: mysql-chain-certs.patch Patch7: mysql-versioning.patch Patch8: mysql-dubious-exports.patch -# Patch9: mysql-disable-test.patch Patch10: mysql-plugin-bool.patch Patch11: mysql-s390-tsc.patch Patch12: mysql-openssl-test.patch Patch13: mysqld-nowatch.patch Patch14: mysql-va-list.patch Patch15: mysql-netdevname.patch +Patch16: mysql-logrotate.patch # RC patch for backports Patch21: mysql-readline.patch @@ -82,6 +84,8 @@ Requires: bash Conflicts: MySQL # mysql-cluster used to be built from this SRPM, but no more Obsoletes: mysql-cluster < 5.1.44 +# Virtual provides present in upstream's RPM (used by some app) +Provides: mysql-client = %{version}-%{release} # When rpm 4.9 is universal, this could be cleaned up: %global __perl_requires %{SOURCE999} @@ -99,6 +103,11 @@ Summary: The shared libraries required for MySQL clients Group: Applications/Databases Requires: /sbin/ldconfig Obsoletes: compat-mysql55 <= %{version} +%if 0%{?rhel} == 5 +# EL-5 mysql 5.0.x have no mysql/mysql-libs +# This circular dep. should make update simpler +Requires: %{name}%{?_isa} = %{version}-%{release} +%endif %description libs The mysql-libs package provides the essential shared libraries for any @@ -219,7 +228,6 @@ rm -f Docs/mysql.info %patch6 -p1 %patch7 -p1 %patch8 -p1 -# si below %patch9 -p1 %patch10 -p1 %patch11 -p1 %if 0%{?fedora} >= 9 || 0%{?rhel} >= 5 @@ -232,6 +240,7 @@ rm -f Docs/mysql.info %endif %patch14 -p1 %patch15 -p1 +%patch16 -p1 # Remi specific patches %patch21 -p1 -b .readline @@ -242,15 +251,12 @@ rm -f mysql-test/t/ssl_8k_key-master.opt # upstream has fallen down badly on symbol versioning, do it ourselves cp %{SOURCE8} libmysql/libmysql.version -# Instead of Pach 9 - Simpler way -cat <>mysql-test/t/disabled.def -# -outfile_loaddata : bug#46895 code wrong, expected results wrong too -sys_vars.plugin_dir_basic : bug#52223 fails for lib64 library directory -innodb.innodb : bug#60155 has platform-dependent results -main.information_schema : fails in mock, ok after install :( -EOF - +# generate a list of tests that fail, but are not disabled by upstream +cat %{SOURCE14} > mysql-test/rh-skipped-tests.list +# disable some tests failing on ARM architectures +%ifarch %{arm} +cat %{SOURCE15} >> mysql-test/rh-skipped-tests.list +%endif %build @@ -351,7 +357,7 @@ cd ../.. LD_LIBRARY_PATH=$PWD/libservices export LD_LIBRARY_PATH -# make test + make test # The cmake build scripts don't provide any simple way to control the # options for mysql-test-run, so ignore the make target and just call it @@ -359,16 +365,18 @@ cd ../.. # --force to continue tests after a failure # no retries please # test SSL with --ssl + # skip tests that are listed in rh-skipped-tests.list # avoid redundant test runs with --binlog-format=mixed # increase timeouts to prevent unwanted failures during mass rebuilds - cd mysql-test ( - # perl ./mysql-test-run.pl --force --retry=0 --mysqld=--binlog-format=mixed --suite-timeout=720 --testcase-timeout=30 - # Run less test to speed up build process - %{__perl} ./mysql-test-run.pl --force --ssl --mysqld=--binlog-format=mixed --suite=main + cd mysql-test + perl ./mysql-test-run.pl --force --retry=0 --ssl \ + --skip-test-list=rh-skipped-tests.list \ + --mysqld=--binlog-format=mixed \ + --suite-timeout=720 --testcase-timeout=30 --suite=main + # cmake build scripts will install the var cruft if left alone :-( + rm -rf var ) - # cmake build scripts will install the var cruft if left alone :-( - rm -rf var %endif %install @@ -472,11 +480,15 @@ rm -f ${RPM_BUILD_ROOT}%{_datadir}/mysql/magic rm -f ${RPM_BUILD_ROOT}%{_datadir}/mysql/ndb-config-2-node.ini rm -f ${RPM_BUILD_ROOT}%{_datadir}/mysql/mysql.server rm -f ${RPM_BUILD_ROOT}%{_datadir}/mysql/mysqld_multi.server -rm -f ${RPM_BUILD_ROOT}%{_datadir}/mysql/mysql-log-rotate rm -f ${RPM_BUILD_ROOT}%{_mandir}/man1/comp_err.1* rm -f ${RPM_BUILD_ROOT}%{_mandir}/man1/mysql-stress-test.pl.1* rm -f ${RPM_BUILD_ROOT}%{_mandir}/man1/mysql-test-run.pl.1* +# put logrotate script where it needs to be +mkdir -p $RPM_BUILD_ROOT%{_sysconfdir}/logrotate.d +mv ${RPM_BUILD_ROOT}%{_datadir}/mysql/mysql-log-rotate $RPM_BUILD_ROOT%{_sysconfdir}/logrotate.d/mysqld +chmod 644 $RPM_BUILD_ROOT%{_sysconfdir}/logrotate.d/mysqld + mkdir -p $RPM_BUILD_ROOT/etc/ld.so.conf.d echo "%{_libdir}/mysql" > $RPM_BUILD_ROOT/etc/ld.so.conf.d/%{name}-%{_arch}.conf @@ -484,6 +496,9 @@ echo "%{_libdir}/mysql" > $RPM_BUILD_ROOT/etc/ld.so.conf.d/%{name}-%{_arch}.conf cp %{SOURCE6} README.mysql-docs cp %{SOURCE7} README.mysql-license +# install the list of skipped tests to be available for user runs +install -m 0644 mysql-test/rh-skipped-tests.list ${RPM_BUILD_ROOT}%{_datadir}/mysql-test + %clean rm -rf $RPM_BUILD_ROOT @@ -730,6 +745,7 @@ fi %attr(0755,mysql,mysql) %dir /var/run/mysqld %attr(0755,mysql,mysql) %dir /var/lib/mysql %attr(0640,mysql,mysql) %config(noreplace) %verify(not md5 size mtime) /var/log/mysqld.log +%config(noreplace) %{_sysconfdir}/logrotate.d/mysqld %files devel %defattr(-,root,root) @@ -764,6 +780,30 @@ fi %{_mandir}/man1/mysql_client_test.1* %changelog +* Tue Feb 12 2012 Remi Collet - 5.5.20-2 +- sync with rawhide +- circular dep for mysql / mysql-libs to fix EL5 update +- provides mysql-client (per user request) + +* Fri Feb 10 2012 Tom Lane 5.5.20-2 +- Revise our test-disabling method to make it possible to disable tests on a + platform-specific basis, and also to get rid of mysql-disable-test.patch, + which broke in just about every upstream update (Honza Horak) +- Disable cycle-counter-dependent regression tests on ARM, since there is + not currently any support for that in Fedora ARM kernels +Resolves: #773116 +- Add some comments to mysqld.service documenting how to customize it +Resolves: #785243 + +* Fri Jan 27 2012 Tom Lane 5.5.20-1 +- Update to MySQL 5.5.20, for various fixes described at + http://dev.mysql.com/doc/refman/5.5/en/news-5-5-20.html + as well as security fixes described at + http://www.oracle.com/technetwork/topics/security/cpujan2012-366304.html +Resolves: #783828 +- Re-include the mysqld logrotate script, now that it's not so bogus +Resolves: #547007 + * Thu Jan 12 2012 Remi Collet - 5.5.20-1 - update to MySQL 5.5.20 Community Server GA http://dev.mysql.com/doc/refman/5.5/en/news-5-5-20.html diff --git a/mysqld.service b/mysqld.service index b3bc486..b7fa095 100644 --- a/mysqld.service +++ b/mysqld.service @@ -1,3 +1,19 @@ +# It's not recommended to modify this file in-place, because it will be +# overwritten during package upgrades. If you want to customize, the +# best way is to create a file "/etc/systemd/system/mysqld.service", +# containing +# .include /lib/systemd/system/mysqld.service +# ...make your changes here... +# For more info about custom unit files, see +# http://fedoraproject.org/wiki/Systemd#How_do_I_customize_a_unit_file.2F_add_a_custom_unit_file.3F + +# For example, if you want to increase mysql's open-files-limit to 10000, +# you need to increase systemd's LimitNOFILE setting, so create a file named +# "/etc/systemd/system/mysqld.service" containing: +# .include /lib/systemd/system/mysqld.service +# [Service] +# LimitNOFILE=10000 + [Unit] Description=MySQL database server After=syslog.target diff --git a/rh-skipped-tests-arm.list b/rh-skipped-tests-arm.list new file mode 100644 index 0000000..a6ae9bf --- /dev/null +++ b/rh-skipped-tests-arm.list @@ -0,0 +1,8 @@ + +# Disable perfschema.func_file_io and perfschema.func_mutex, which fail +# because cycle counter returns 0 every time on ARM architectures. +# This is caused by missing hardware performance counter support on ARM. +# Discussion about fixing that can be found in RH bug #741325. + +perfschema.func_file_io : rhbz#773116 cycle counter does not work on arm +perfschema.func_mutex : rhbz#773116 cycle counter does not work on arm diff --git a/rh-skipped-tests-base.list b/rh-skipped-tests-base.list new file mode 100644 index 0000000..a855f78 --- /dev/null +++ b/rh-skipped-tests-base.list @@ -0,0 +1,20 @@ +# Disable the outfile_loaddata test, which as of 5.1.38 is giving +# platform-dependent results, with the "expected" results being arguably the +# wrong ones. This is upstream at http://bugs.mysql.com/bug.php?id=46895 +# (note that upstream has also disabled it, but only for Solaris, so we still +# need to disable it here). +# Still broken in 5.5.14, despite alleged fix. + +outfile_loaddata : bug#46895 code wrong, expected results wrong too + +# Disable sys_vars.plugin_dir_basic, which is broken because +# mysql-test-run doesn't set the right value of MYSQL_LIBDIR. +# Upstream at http://bugs.mysql.com/bug.php?id=52223 +# Still fails in 5.5.16, even though upstream marked it closed. + +sys_vars.plugin_dir_basic : bug#52223 fails for lib64 library directory + +# Disable innodb.innodb, which is showing platform-dependent results +# as of 5.5.9. Upstream at http://bugs.mysql.com/bug.php?id=60155 + +innodb.innodb : bug#60155 has platform-dependent results -- cgit