On Fri, Apr 01, 2022 at 10:21:50AM -0700, Andres Freund wrote:
> right now I am looking at a test added in the shmstats patch that's slow on
> CI, on windows only. Unfortunately the regress_log_* output is useless as-is
> to figure out where things hang.
>
> I've hit this several times before. Of course it's not too hard to hack up
> something printing elapsed time. But ISTM that it'd be better if we were able
> to prefix the logging into regress_log_* with something like
> [timestamp + time since start of test]
> or
> [timestamp + time since start of test + time since last log message]
>
>
> This isn't just useful to figure out what parts of test are slow, but also
> helps correlate server logs with the regress_log_* output. Which right now is
> hard and inaccurate, requiring manually correlating statements between server
> log and the tap test (often there's no logging for statements in the
> regress_log_*).
+1
--
Nathan Bossart
Amazon Web Services: https://aws.amazon.com