Skip to content

Add regression suite and source-aware legends - #9

Merged
bd4rex merged 5 commits into
mainfrom
agent/regression-test-suite
Aug 9, 2026
Merged

bd4rex merged 5 commits into
mainfrom
agent/regression-test-suite

Conversation

@bd4rex

@bd4rex bd4rex commented Aug 7, 2026 •

Copy link
Copy Markdown
Owner

What changed

  • add a machine-readable catalog of 16 regression cases derived from the project's iterative map, resource, data, performance, and recovery work
  • add one PowerShell entry point with static, browser, full, and recovery profiles
  • add repository contracts for JSON, PowerShell syntax, bilingual document parity, local Markdown links, test metadata, performance baselines, and Playwright-image completeness
  • run the static profile automatically for pull requests and pushes to main
  • package the performance test and its three-run baseline in the Playwright image
  • normalize the weather JQ program to LF before Linux-container execution
  • preserve and test consistent 台湾省 naming across package prompts, resource names, feature details, and world-overview labels
  • make the map legend source-aware for local vector, OSM Carto, global overview, OpenFreeMap, and active overlay layers
  • add detailed, expandable road, land-use, facility, label, and overview legend groups with browser regression coverage
  • document usage, prerequisites, side effects, evidence, and maintenance rules in English and Chinese

Why

The repository had strong individual smoke scripts, but no single development workflow or durable catalog connecting past regressions to the checks that prevent them. The documented performance command also could not run from the UI-test image because the script and baseline were not copied into it.

Running the new full suite additionally exposed repeated weather synchronization failures for newly installed country packages. The Windows checkout supplied a CRLF JQ program to a Debian container, where JQ rejected the carriage return before processing the first uncached country.

The previous fixed legend also did not explain the symbols of the map source currently on screen. It now follows source and layer state, with compact defaults and expandable details.

Impact

Developers can now choose a test cost appropriate to each change, while GitHub catches repository and documentation regressions without requiring the private local GIS payload. Runtime and recovery profiles remain local because they depend on the prepared eight-service stack and offline products.

Map users now get an accurate legend when switching between detailed offline/online sources and the low-zoom world overview, including currently enabled overlays.

Validation

  • static: 6 JSON files, 47 PowerShell scripts, 21 bilingual pairs, 194 local Markdown links, 16 test cases, 3 performance samples, and 4 browser tests
  • eight-service health check: passed
  • API, regional-resource, CRUD/versioning, search/routing, export, and owned-media lifecycle: passed
  • main UI Playwright: passed, including vector and OSM Carto detailed legend assertions
  • resource-console Playwright: passed
  • world-map Playwright: passed, including overview, online OSM, and OpenFreeMap legend assertions
  • performance Playwright: passed with DOM 267 ms, map canvas 1100 ms, system ready 1616 ms, and one map-pack request
  • complete full profile: passed in 423.8 seconds before the legend-only follow-up
  • Node syntax checks for the two changed browser tests and app script: passed
  • git diff --cached --check: passed

The isolated recovery profile was added but not run for this change because no recovery payload or recovery implementation changed.

@bd4rex bd4rex changed the title Add regression test suite Add regression suite and source-aware legends Aug 7, 2026
@bd4rex
bd4rex merged commit 4d76133 into main Aug 9, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant