Tests fail _sometimes_

[eulerto]

Tests fail _sometimes_

[eulerto]

@afiskon Is it version 10?

[afiskon]

Yes, it's 10.

[eulerto]

@afiskon I couldn't reproduce it. Could you check again? I pushed some PR and did some enhancements.

[afiskon]

@eulerto This particular issue doesn't reproduce anymore. However now the message test always fails like this:

https://afiskon.ru/s/8b/155998a5fb_regression.diffs.txt
https://afiskon.ru/s/7a/c09baf7e51_regression.out.txt

Reproduces on both PostgreSQL 10 and 11.

[afiskon]

Here is how exactly I build and run PostgreSQL:

https://afiskon.ru/s/47/d46b9b783e_quick-build.sh.txt
https://afiskon.ru/s/fa/e2b93ea9c4_single-install.sh.txt

Also I manually install contrib/test_decoding.

[eulerto]

@afiskon What is your server encoding?Could you check regression tests under locale C or any UTF-8 (if it is not the case)?

[afiskon]

Hello,

I'm using locale en_US.UTF-8. I execute regression tests for the master branch (i.e. 11) a few times a day, they are fine.

The environment is Arch Linux x64 if that matters, pretty much default configuration with i3wm.

[rich617]

These regression tests still fail intermittently on Windows with the same type of diff where there are one or more empty list entries extra from the test run. This is over two years in. Does anyone know where these extra entries are coming from? Can we modify the tests to filter out the empty entries from the test output?