Skip to content

TESTING.md

Latest commit

 

History

History
112 lines (79 loc) · 3.81 KB

TESTING.md

File metadata and controls

112 lines (79 loc) · 3.81 KB
Raw

TESTING

unit tests

the matcher state machine and override parsing are host-testable without the VM:

$ gjs -m tests/matcher_test.js
$ gjs -m tests/common_test.js

preflight

before committing to a full ~8 minute suite, run the fast environment check:

$ scripts/vm-test.sh preflight

it verifies the window-control instrument responds, the test monitors are settable to the single-monitor baseline at a real advertised mode, and the extension comes up. the same check runs automatically as a session-scoped fixture (verify_environment), so a broken environment aborts the run in seconds with an unmistakable ENVIRONMENT NOT READY message instead of surfacing minutes later as a cluster of failures that look like regressions.

reliability

the harness separates environment/harness breakage from product regressions so an environment glitch never stalls development:

  • monitor/display operations raise HarnessError; pytest labels these HARNESS/ENVIRONMENT PROBLEM, distinct from an assertion (a real regression). scripts/vm-test.sh preflight diagnoses them.
  • set_monitors selects a real advertised mode per connector (refresh rates differ between the virtual outputs) rather than assuming one global mode.
  • each test restores the single-monitor baseline during setup, so a monitor test that fails mid-reconfiguration cannot cascade into the next test.

bootstrap

should be run after extension code changes

  • scripts/vm-test.sh uninstall
  • scripts/vm-test.sh reboot
  • scripts/vm-test.sh install
  • scripts/vm-test.sh logout (fresh installs are not discovered until the session restarts)

test-only changes need no bootstrap: the tests directory is mounted into the guest at /srv/smart-auto-move and conftest verifies the installed extension matches the source by hash.

the VM is started automatically when needed (scripts/vm-test.sh start).

cleanup

should be run between each test

  • disable the extension
  • reset the dconf settings
  • restore the single-monitor baseline (only if a prior test left it changed)
  • enable the extension

orchestration

window orchestration is done by /srv/window-control/windowbot.py

test configs for windowbot are in /srv/window-control/testdata/

some configs have long delays with title changes/etc. tests should be resilient to those.

running stories

$ scripts/vm-test.sh pytest                                    # full suite
$ scripts/vm-test.sh pytest test_story_12_title_migration_early.py

stories

default settings, calculator restore

  • open calculator, move, resize, close, reopen (should restore)

sync mode ignore, override calculator restore

  • set default sync mode to IGNORE
  • open calculator, move, close, reopen (should NOT restore)
  • add an app override for calculator (mode RESTORE, 0.7 threshold)
  • open calculator, move, close, reopen (should restore)
  • open nautilus, move, close, reopen (should NOT restore)

workspace and tiling

  • launch the "fasttitle" windowbot config
  • wait for the window titles to settle
  • move the windows to the workspace and tile as described by their titles
  • quit and restart the windowbot session with "fasttitle" (should restore)
  • quit and restart the windowbot session with "slowtitle" (should restore, with long delays)

secondary monitors

  • launch the "single" windowbot config
  • tile the window to the right on the primary monitor
  • enable the secondary monitor
  • move the window to the secondary monitor and tile to the left
  • disable the secondary monitor (should move to primary and tile right)
  • enable the secondary monitor (should move to secondary and tile left)
  • quit and restart the windowbot session with "single" (should restore to secondary, tile left)
  • quit the windowbot session
  • disable the secondary monitor
  • start the windowbot session with "single" (should restore to primary, tile right)