try to parallelize test jobs

[lunaticusgreen]

No description provided.

[coveralls]

Pull Request Test Coverage Report for Build 22716597308

Details

• 0 of 0 changed or added relevant lines in 0 files are covered.
• 695 unchanged lines in 9 files lost coverage.
• Overall coverage decreased (-22.1%) to 40.141%

---

Files with Coverage Reduction	New Missed Lines	%
pkg/requests/request_response.go	4	0.0%
pkg/main.go	5	0.0%
pkg/query.go	54	0.0%
pkg/eval/eval_query.go	59	76.56%
pkg/resource_handlers.go	59	0.0%
pkg/client.go	77	0.0%
pkg/datasource.go	85	0.0%
pkg/response.go	161	0.0%
pkg/parser.go	191	0.0%

Totals
Change from base Build 22607053745:	-22.1%
Covered Lines:	1476
Relevant Lines:	3677

---

💛 - Coveralls

[Selfeer]

You don't really need a custom solution like should_run() and suite, instead you can just rely on the --only flag, even with your matrix approach to have something like:

- group: window-functions
  only: "*/window functions/*"
- group: legacy-alerts
  only: "*/legacy alerts/*"

-----------------------------------------

run: cd ./tests/testflows/ &&
  python3 -u ./regression.py
  --before=0.1 --after=0.1
  --only "${{ matrix.only }}"
  --log raw.log -o new-fails

But generally we've moved to using a reusable-suite for most our projects:

https://github.com/Altinity/clickhouse-regression/blob/main/.github/workflows/reusable-suite.yml

These are the most basic comments that I have that can make you approach cleaner, but generally I would just point an AI agent to our clickhouse-regression suite and make it take our approach. right now your regression.py looks a bit messy.

Also as a note: keep -o new-fails - The PR downgrades to -o short, losing the most useful CI output format.

[Slach]

why you remove it?

[Slach]

why you remove it?