ci: enforce SQLAlchemy 2.0 deprecation warnings in unit-test CI

[rusackas]

SUMMARY

Part of the SQLAlchemy 2.0 migration battleplan: #40273

pytest.ini's filterwarnings already promotes several SQLAlchemy 2.0 deprecation-warning categories from a bare warning to a hard error, as each gets fixed and locked in (see #41914 for 3 more). But SQLALCHEMY_WARN_20 was never actually set anywhere in CI — grepped the whole repo, it only appears in a comment in pytest.ini. That means those RemovedIn20Warning categories never actually fire during a normal CI run, so the "enforcement" was a no-op: a regression on any already-"fixed" category would pass CI silently, and a contributor would only catch it if they remembered to set the env var locally.

This sets SQLALCHEMY_WARN_20: "1" on the unit-tests job so the categories already promoted to error: in pytest.ini are actually exercised on every PR and push.

TESTING INSTRUCTIONS

Ran the exact CI command locally against master's current pytest.ini (3 promoted categories):

SQLALCHEMY_WARN_20=1 SUPERSET_TESTENV=true SUPERSET_SECRET_KEY=not-a-secret \
  pytest --durations-min=0.5 --cov-report= --cov=superset ./tests/common ./tests/unit_tests --cache-clear --maxfail=50

Only 2 pre-existing, unrelated flaky tests fail (test_get_time_filter ×6, test_previous_calendar_quarter), and both reproduce identically with no env vars set at all — confirmed unrelated to this change (date-boundary-sensitive).

ADDITIONAL INFORMATION

• Has associated issue:
• Required feature flags:
• Changes UI
• Includes DB Migration
• Introduces new feature or API
• Removes existing feature or API

[bito-code-review]

Bito Automatic Review Skipped - Files Excluded

Bito didn't auto-review this change because all changed files are in the exclusion list for automatic reviews. No action is needed if you didn't intend for the agent to review it. Otherwise, to manually trigger a review, type /review in a comment and save.
You can change the excluded files settings here, or contact your Bito workspace admin at evan@preset.io.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 64.56%. Comparing base (1fd43ff) to head (4299573).
⚠️ Report is 2 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #41915      +/-   ##
==========================================
- Coverage   64.57%   64.56%   -0.02%     
==========================================
  Files        2713     2718       +5     
  Lines      151269   151491     +222     
  Branches    34793    34802       +9     
==========================================
+ Hits        97684    97808     +124     
- Misses      51760    51854      +94     
- Partials     1825     1829       +4     

Flag	Coverage Δ
hive	39.07% <ø> (+0.06%)	⬆️
mysql	57.44% <ø> (+<0.01%)	⬆️
postgres	57.50% <ø> (-0.01%)	⬇️
presto	40.58% <ø> (+0.05%)	⬆️
python	58.90% <ø> (-0.01%)	⬇️
sqlite	57.09% <ø> (+<0.01%)	⬆️
unit	100.00% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[hy144328]

I think this would be great to catch regression in the migration.
I considered it in #40274 but I was reluctant to make the step at the time.
Currently, the unit tests would product 45131 lines locally with warnings turned on with 45218 warnings produced.
I was not sure whether this would overwhelm the CI, a log database, etc.
Other than that, my two cents is that this is not a controversial change.

==== 7 failed, 9756 passed, 5 skipped, 45218 warnings in 782.83s (0:13:02) =====

[rusackas]

@hy144328 thanks for the numbers. CI ran green here so the volume did not seem to overwhelm anything, and filterwarnings only hard-errors the few categories we have actually migrated, so the rest stay print-only. Curious what your 7 local failures were though, if they fall outside that set?

[hainenber]

LGTM. Let's do this!

[hy144328]

@hy144328 thanks for the numbers. CI ran green here so the volume did not seem to overwhelm anything, and filterwarnings only hard-errors the few categories we have actually migrated, so the rest stay print-only. Curious what your 7 local failures were though, if they fall outside that set?

@rusackas Sorry for the late reply.

I have been noticing errors locally here and there that did not show up in the CI.
So I mostly ignored them to keep things moving.
I have just noticed that they all go away when I upgrade my virtual environment from Python 3.10 to 3.11.
For instance, there is the LalitMaganti/syntaqlite dependency, which claims to be Python 3.10 compatible.
But it causes unit tests to fail because it imports enum.StrEnum , which has only been introduced in Python 3.11 .
I wonder whether it is worthwhile raising a ticket for this.

[rusackas]

I actually opened a revert PR for that dependency today, but I'm considering removing support for python 3.10 instead. That might be more forward leaning, but I want to get more consensus.