[Maria-developers] New BuildBot tests - compiling the MySQL drivers for PHP and Perl
Hello, I have installed a new builder "compile-connectors", which will compile the DBD::mysql and the PHP (5.3) mysql connector (the oldest of them all). The purpose is to make sure that MariaDB does not break compatibility with MySQL when it comes to things such as headers, file locations, compilability, linking and loadability of the compiled module. I will try to get some more things tested, such as the newer PHP driver, the C++ connector and drivers containing the embedded server. If you can think of any other connectors or important applications that link or load libmysql, let me know and I will try to add them too. I do not think the new test will catch bugs such as LP#674812, which showed up when mixing and matching OS packages, downloaded packages and self-compiled stuff, but we can at least catch the regression if our headers get modified enough for the third-party drivers to become uncompilable. My next step will be to automate the testing of the Windows installer and the installation of MariaDB as a service. Any thoughts you may have in that area are also appreciated. Thank you. Philip Stoev
"Philip Stoev" <pstoev@askmonty.org> writes:
I have installed a new builder "compile-connectors", which will compile the DBD::mysql and the PHP (5.3) mysql connector (the oldest of them all). The purpose is to make sure that MariaDB does not break compatibility with MySQL when it comes to things such as headers, file locations, compilability, linking and loadability of the compiled module.
Oh, that's excellent!
I will try to get some more things tested, such as the newer PHP driver, the C++ connector and drivers containing the embedded server. If you can think of any other connectors or important applications that link or load libmysql, let me know and I will try to add them too.
The JDBC drivers is what springs to my mind ...
I do not think the new test will catch bugs such as LP#674812, which showed up when mixing and matching OS packages, downloaded packages and self-compiled stuff, but we can at least catch the regression if our headers get modified enough for the third-party drivers to become uncompilable.
Agree. My take is that users are in general somewhat forgiving about such cornercase bugs, especially if we react timely to failure reports. On the other hand, total breakage, like packages that are uninstallable under any conditions etc., that is harder to forgive ("didn't they test this at all?"). So I think the automated tests have great value, even if they mostly test the basic stuff. - Kristian.
participants (2)
-
Kristian Nielsen -
Philip Stoev