[Maria-developers] FYI: 10.3 master branch head builds OK Launchpad all architectures
Hello! Just for the record, I've done some small tweaks to 10.3 so that it now builds perfectly and passes build tests on all Launchpad architectures that I have access to. https://launchpad.net/~mysql-ubuntu/+archive/ubuntu/mariadb-10.3/+build s?build_text=&build_state=all Let's hope it stays like this for the rest of the current 10.3 cycle and will not regress by before it is announced GA.. On buildbot the Debian sid and Ubuntu Bionic builders pass now. There are however some sporadic failures left. http://buildbot.askmonty.org/buildbot/grid?branch=10.3&category=main&ca tegory=experimental On Travis everything else passes apart from one test, which has been consistently failing for some time. Everything was green in January when I wrote the protected branches blog post, so this one is a consistently failing regression. https://travis-ci.org/MariaDB/server/builds/362466962 - Otto
On 05/04/18 19:48, Otto Kekäläinen wrote:
Hello!
Just for the record, I've done some small tweaks to 10.3
Great work so far.
so that it now builds perfectly and passes build tests on all Launchpad architectures that I have access to.
Needs a bit more than build tests: The use of /etc/mysql/mariadb.conf.d/ is admirable however it appears its never read (MDEV-15627). This path needs to be integrated into the build system and the executables not just a debian move files around implementation. Is the reading of /etc/mysql/my.cnf / /etc/mysql/my.conf.d desired? This path also isn't used by the system mariadb@ multi-instance service file (support-files/mariadb@.service.in). The rather ugly /etc/mysql/conf.d/my%I.cnf is (sorry). A small variant to pr #510 could set MYSQLD_MULTI_INSTANCE=--defaults-file=/etc/mysql/mariadb.%I.cnf which a consistent way using the alternative configuration directories. Is there a more preferred way? Log-rotation is probably broken too afaik pid files may not exists - solution in MDEV-13050 however check both conditions if necessary. Why is /usr/include/mariadb/server/private/ installed? AFAIK there is no supported API/ABI here. /etc/mysql/conf.d/mysqld_safe_syslog.cnf assuming this is debian/additions/mysqld_safe_syslog.cnf and mysqld_safe isn't used in the a systemd environment it can be removed. debian/additions/mariadb.cnf - seems to encourage changing character set here. How about this occurs in a /etc/mysql/mariadb.conf.d/charset.cnf file? (also note header). debian-start, MDEV-14622, is also a problem.
..in continuation to the 10.3 overview, I made (by hand) made this table which includes the names of the failing tests in question. https://imgur.com/a/dxcJk I built the exact same commit built 4 times in a row with no changes, so all tests that fail are sporadic. This gives a nice overview of which tests and how often they fail. Most of them seems to be Galera related, but there are also ones about RocksDB and versioning.* and InnoDB. I encourage everybody on this mailing list to pay attention to these. We really need to get rid of these test failures now when 10.3 is supposed to be GA very soon. - Otto
Regarding the test that permanently fails on Travis-CI and keeps it from being green I filed https://jira.mariadb.org/browse/MDEV-15814 If Travis-CI was all green (=this one test fixed or disabled) we could perhaps enable protected branches feature on the 10.3 branch to protect it from not breaking down anymore this late in the release cycle. What do you think? https://mariadb.org/protected-branches-ensuring-code-quality-git/ - Otto
On 10/04/18 02:04, Otto Kekäläinen wrote:
Regarding the test that permanently fails on Travis-CI and keeps it from being green I filed https://jira.mariadb.org/browse/MDEV-15814
If Travis-CI was all green (=this one test fixed or disabled) we could
PR with commit to disable test from Dec 1, 2017: https://github.com/MariaDB/server/pull/505/commits/9fce41be75a1620f11bdcbfd3...
Did a new build today and there is a test failure like this now: main.myisampack w3 [ fail ] Test ended at 2018-04-10 09:02:35 CURRENT_TEST: main.myisampack --- /<<PKGBUILDDIR>>/mysql-test/main/myisampack.result 2018-03- 31 19:34:34.000000000 +0000 +++ /<<PKGBUILDDIR>>/mysql-test/main/myisampack.reject 2018-04- 10 09:02:34.965478238 +0000 @@ -113,7 +113,7 @@ # ===== myisampack.4 ===== #Tests the myisampack join operation with an existing destination .frm,.MYI,.MDI #the command should fail with exit status 2 -myisampack: Can't create/write to file (Errcode: 17 "File exists") +myisampack: Can't create/write to file (Errcode: 17 "File exists) Aborted: file is not compressed DROP TABLE t1,t2,t3; DROP TABLE mysql_db1.t1; mysqltest: Result length mismatch - saving '/<<PKGBUILDDIR>>/builddir/mysql- test/var/3/log/main.myisampack/' to '/<<PKGBUILDDIR>>/builddir/mysql- test/var/log/main.myisampack/' i386 builds, log at https://launchpadlibrarian.net/364229537/buildlog_ubuntu-bionic-i386.ma riadb- 10.3_1%3A10.3.0~ubuntu18.04.1~1523349076.7980364804f_BUILDING.txt.gz Interestingly one " is missing.
participants (2)
-
Daniel Black -
Otto Kekäläinen