On Tue, Jun 07, 2022 at 12:18:52PM +0100, Dagfinn Ilmari Mannsåker wrote:
> Point, fixed in the attached v2. I've also added a test that truncation
> and disabling works.
The tests are structured so as all the queries are run first, then the
full set of logs is slurped and scanned. With things like tests for
truncation, it becomes easier to have a given test overlap with
another one in terms of pattern matching, so we could silently lose
coverage. Wouldn't it be better to reorganize things so we save the
current size of the log file (as of -s $node->logfile), run one query,
and then slurp the log file based on the position saved previously?
The GUC updates had better be localized in each sub-section of the
tests, as well.
--
Michael