Details
Trivial
Description
Before 5.5 (or possibly before MDEV-30), skip-combinations provided on MTR command line worked as the name suggests. In 5.5 it is declared and accepted, but ignored (and in the code seems to be no use of it, instead there is a hash which is read from the suite file).
MariaDB 5.3:
ubuntu11:~/maria-5.3/mysql-test$ perl ./mtr binlog.binlog_tmp_table --skip-combinations ============================================================================== TEST RESULT TIME (ms) ------------------------------------------------------------ worker[1] Using MTR_BUILD_THREAD 300, with reserved ports 16000..16019 binlog.binlog_tmp_table [ pass ] 2055 ------------------------------------------------------------ The servers were restarted 0 times Spent 2.055 of 4 seconds executing testcases Completed: All 1 tests were successful.
MariaDB 5.5 (revno 3462):
perl ./mtr binlog.binlog_tmp_table --skip-combinations ============================================================================== TEST RESULT TIME (ms) or COMMENT -------------------------------------------------------------------------- worker[1] Using MTR_BUILD_THREAD 300, with reserved ports 16000..16019 binlog.binlog_tmp_table 'mix' [ pass ] 326 binlog.binlog_tmp_table 'row' [ skipped ] Neither MIXED nor STATEMENT binlog format binlog.binlog_tmp_table 'stmt' [ pass ] 323 -------------------------------------------------------------------------- The servers were restarted 2 times Spent 0.649 of 3 seconds executing testcases Completed: All 2 tests were successful. 1 tests were skipped, 1 by the test itself.
It's a good option, would be nice to have it back if possible.
Gliffy Diagrams
Attachments
Activity
- All
- Comments
- Work Log
- History
- Activity
- Transitions
I expected it to be as simple as completely ignoring the combinations file and let the user or fate decide whether it contained mandatory parameters or optional ones. Like in the example above, ignore 3 binlog formats and run only once, with whatever server or MTR sets by default.
That said, I don't insist on it, since it can be achieved by other means (hence the 'Trivial' priority). If it was intentional, lets just remove the option from usage printed by MTR, and from the list of options it reads and saves. Then, if somebody puts the option on the command line, it will be passed through to the server, and server will die of ignorance – which is also a bit confusing, but not as much as a declared but non-functional option, and the behavior will be consistent with other skip-<don't know what> options.
I've intentionally removed it in
MDEV-30. In MySQL combinations were, more or less, optional. One could imagine running tests without them. AfterMDEV-30they're essential and infuence heavily what tests are run and how. If combinations were completely ignored, the server, probably, wouldn't even start.What exactly do you want to achieve with --skip-combinations option?