Re: On login trigger: take three - Mailing list pgsql-hackers

Hi Pavel,

Thanks for pointing out the tests. I completely agree that using an exception inside on-login trigger should be tested. It cannot be done via regular *.sql/*.out regress tests, thus I have added another perl test to authentication group doing this.

Attached v36 patch contains this test along with the fresh rebase on master.

--

 best regards,
    Mikhail A. Gribkov

e-mail: youzhick@gmail.com

http://www.flickr.com/photos/youzhick/albums

http://www.strava.com/athletes/5085772
phone: +7(916)604-71-12
Telegram: @youzhick

On Thu, Jan 12, 2023 at 9:51 AM Pavel Stehule <pavel.stehule@gmail.com> wrote:

Hi

I checked this patch and it looks well. All tests passed. Together with https://commitfest.postgresql.org/41/4013/ it can be a good feature.

I re-tested impact on performance and for the worst case looks like less than 1% (0.8%). I think it is acceptable. Tested pgbench scenario "SELECT 1"

pgbench -f ~/test.sql -C -c 3 -j 5 -T 100 -P10 postgres

733 tps (master), 727 tps (patched).

I think raising an exception inside should be better tested - not it is only in 001_stream_rep.pl - generally more tests are welcome - there are no tested handling exceptions.

Regards

Pavel