fix: regenerate state configs from dev; table-view wait matched an icon, not the table

[saqibmanan]

Part of #45 (get the pre-merge full regression green). Refs #44.

Why

Every automatic run targets dev: the PR gate, main.yml (pointed at dev until the prod move) and nightly. The committed config/states/*.yaml were discovered on 2026-06-17 and no longer match dev. The multistate tests select each configured indicator by its exact name, so every renamed or moved indicator fails its whole section.

A fresh baseline on main (Dispatch Tests, regression on dev, run 36150264203) failed on exactly these names:

• Assam exposure / gov response: Elderly population, Children population (moved to vulnerability on dev), and the six Total Value of Flood Tenders… indicators (renamed to Flood Tenders : …)
• Odisha exposure / vulnerability: Population*, Total Number of Households*, …flood and cyclone names

Change

Regenerated from dev with scripts/discover_state_indicators.py (2026-09-25), one commit per state. Himachal Pradesh is unchanged; its only diff was the timestamp.

State	committed	dev
Assam	30	32
Bihar	7	15
Odisha	19	22
Uttar Pradesh	6	22
Himachal Pradesh	32	32
Total	94	123

Evidence

• Before, on main in CI: test_section_all_indicators_map_view[assam-Assam-exposure] failed with "2/4 indicators FAILED: Elderly population, Children population", and [assam-Assam-government-response] with "6/6 indicators FAILED".
• After, locally against dev with this branch's configs: the same two tests give 2 passed.
• This PR's own Full Regression (pre-merge) run is the full check.

Prod

Prod (drr.open-contracting.in) still serves Assam only: the other four states 404 on 2026-09-25, so #43 holds and the prod move hasn't happened. Prod isn't targeted by any automatic run, so these dev-based configs don't break anything there. When prod catches up with dev they will already match it. Same plan as before, with no environment-specific config added.

Second fix: every table-view test was failing (test bug)

Baseline shard 3: 15 of 15 TestMultiStateIndicatorsTableView tests failed in every state, including Himachal Pradesh, which has no drift. Each indicator failed with Table validation failed: Message: (a blank timeout).

The cause: the wait used EC.visibility_of_element_located((By.CSS_SELECTOR, "table, .table, [class*='table']")). Selenium takes the first match in the document, and that first match is a Tabler icon (<svg class="tabler-icon …">, since "tabler" contains "table"), not the table. Waiting on that icon's visibility timed out while the real table was rendered. I checked in the browser with the test's exact flow (Assam → Table View → Sivasagar / Sibsagar → Total Number of Households): a <table> with the right column showed within 1 ms.

The fix waits for any displayed <table>.

• Red, locally with --reruns off: [uttar-pradesh-Uttar pradesh-hazard] → 1 failed, "Table validation failed" on both indicators
• Green, after the fix: [uttar-pradesh-Uttar pradesh-hazard] and [assam-Assam-exposure] → 2 passed

The other two shard 3 failures (TestSectionCoverageByState Odisha exposure, Assam government response: "No indicators successfully tested") are the same indicator-name drift as above.

Third change: data gaps are reported, not failed

The first rerun of this PR (36165577942) left 4 failures, all of them the app working correctly with nothing to show:

• Table view shows "No data available." HP vulnerability (the four Vulnerability Index indicators, sanitation and one more), Odisha "…Affected by Flood and Cyclone" (2), Assam "Immediate Measure Flood Tenders : Total Value". For HP Social Vulnerability Index I checked by hand: no data for either May or June 2026, at state or district level.
• Chart view doesn't offer 2 UP government-response indicators ("Flood Tenders Under SDRF : Total value", "State Disaster Relief Funds Allocated District-wise During SEC meetings : Total Value"), while map and table view list all 8.

Decision (agreed): these are reported, not failed.

• AnalyticsPage.is_no_data_displayed() and is_indicator_listed().
• Table view: an explicit "No data available." → warnings.warn.
• Chart view: an indicator the view doesn't list → warnings.warn. Map and table view still fail if an indicator really disappears.
• pytest.ini: dropped --disable-warnings so the warnings summary is actually printed. The only other warnings in the suite were my own local -p no:rerunfailures config warnings.

Verified locally against dev: UP chart view plus the Assam and Odisha table views → 3 passed. The UP run's summary names exactly the 2 expected indicators.

A data card listing every "No data available." indicator will follow from this run's warnings summary.

Follow-up (not in this PR)

The chart and map waits (canvas, svg, .chart, [class*='chart'] and canvas, svg) are satisfied by any visible SVG icon, so they pass even if no chart or map renders. That's the same substring-selector trap as the table wait, in the other direction. It needs the real chart and map containers; to be tracked as a separate card.

Baseline summary (main, run 36150264203)

• Shards 4–8 passed.
• Shards 1 and 2: 9 failures. 8 are name drift; 1 is Assam hazard table view (the icon bug).
• Shard 3: 17 failures. 15 are table view (the icon bug); 2 are section coverage (drift).

[github-actions]

Test Results (all)

368 tests  +6   341 ✅ +40   2h 16m 31s ⏱️ - 3h 57m 59s
  8 suites ±0    27 💤  -  8 
  8 files   ±0     0 ❌  - 26 

Results for commit 90dad34. ± Comparison against base commit d0b99a2.

♻️ This comment has been updated with latest results.

[github-actions]

QA Tests ✅ passed

• Markers: all
• Sharding: 8 shards × 2 workers
• Target env: dev

Download the consolidated-test-report artifact for the merged report.html.