Two pre-existing issues exposed by running tests in CI for the first time:

1. jasmine-sinon@0.4.0 calls jasmine.pp() which was removed as a global
   in jasmine 4+. jasmine-core was bumped to 5.1.1 in #40956 without
   this being caught. Add a jasmine.pp shim in specHelper.js that
   redirects to JSON.stringify.

2. setupchecksSpec.js checkDataProtected tests were changed from
   requests[0].respond() to respondWith+respond() in cf22336 to fix
   a jasmine 5.x issue, but this breaks with nise 6.1.x (sinon 19+) due
   to nise behavior changes. The checkWebDAV tests use requests[0].respond()
   successfully, so revert checkDataProtected to the same direct-respond
   pattern.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Thomas Müller <1005065+DeepDiver1975@users.noreply.github.com>

* ci: add JavaScript unit test job to GitHub Actions

Mirror the Drone CI javascript() pipeline in GitHub Actions by adding a
local reusable workflow (js-unit.yml) that runs `make test-js` via Karma,
and wiring it into ci.yml.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Thomas Müller <1005065+DeepDiver1975@users.noreply.github.com>

* ci(js-unit): install Firefox from Mozilla PPA instead of snap

On ubuntu-latest (24.04), apt-get install firefox installs a snap
wrapper. Snap-sandboxed Firefox cannot connect to Karma's loopback
socket (localhost:9876), causing all 3 launch attempts to time out.

Install the native deb from the Mozilla PPA with a pin rule that
ensures it wins over the snap wrapper on any future apt-get run.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Thomas Müller <1005065+DeepDiver1975@users.noreply.github.com>

* fix(js-tests): pin sinon to ^19.0.5 to restore fakeServer in browser

sinon v20.0.0 removed nise (the XHR simulation library) from its
browser UMD bundle (pkg/sinon.js), causing sinon.fakeServer to be
undefined. karma-jasmine-sinon injects pkg/sinon.js and relies on
window.sinon.fakeServer, so every test using fakeServer.create() in
specHelper.js fails.

sinon v19.0.5 is the last version whose browser bundle still exports
fakeServer. The broken bump to ^22.0.0 was introduced in commit
aaf5f0d (dependabot bump #41548) without test validation.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Thomas Müller <1005065+DeepDiver1975@users.noreply.github.com>

* fix(js-tests): fix jasmine 5.x incompatibilities in test helpers

Two pre-existing issues exposed by running tests in CI for the first time:

1. jasmine-sinon@0.4.0 calls jasmine.pp() which was removed as a global
   in jasmine 4+. jasmine-core was bumped to 5.1.1 in #40956 without
   this being caught. Add a jasmine.pp shim in specHelper.js that
   redirects to JSON.stringify.

2. setupchecksSpec.js checkDataProtected tests were changed from
   requests[0].respond() to respondWith+respond() in cf22336 to fix
   a jasmine 5.x issue, but this breaks with nise 6.1.x (sinon 19+) due
   to nise behavior changes. The checkWebDAV tests use requests[0].respond()
   successfully, so revert checkDataProtected to the same direct-respond
   pattern.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Thomas Müller <1005065+DeepDiver1975@users.noreply.github.com>

* fix(js-tests): restore test isolation broken by Jasmine 5 random ordering

Two tests relied on shared mutable state that leaked between specs when
Jasmine 5 randomized execution order:

- mimeTypeSpec: afterEach restored OC.currentTheme.name but not
  OC.currentTheme.directory, causing shareSpec's OC.imagePath() calls to
  return the themed '/themes/abc/...' path instead of the expected default.

- setupchecksSpec: oc_dataURL was set once at describe-body level and a
  test mutated it to false with no cleanup, so subsequent tests saw
  oc_dataURL===false and checkDataProtected() returned early without
  making any XHR, leaving suite.server.requests[0] undefined.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Thomas Müller <1005065+DeepDiver1975@users.noreply.github.com>

* fix(js-tests): fix three more test isolation issues from Jasmine 5 ordering

Three more leaks exposed by Jasmine 5's randomised test execution:

1. mimeTypeSpec: afterEach restored OC.currentTheme name/directory but
   did not flush OC.MimeType._mimeTypeIcons. Tests running after the theme
   describe got a cache hit with the themed path for 'dir' and failed.

2. l10nSpec: 'calls callback if locale is en' stubbed OC.getLocale but
   never called localeStub.restore(), leaving getLocale wrapped for any
   test that ran afterward and tried to stub it again.

3. appSpec (files_sharing): app.js registers a notification handler via
   $(document).ready at load time. The 'Action events' beforeEach called
   registerNotificationHandler() a second time without removing the first,
   so each body event fired two handlers and reload counts were 2 instead
   of 1. Fix: clear all OCA.Notification.Action handlers before adding the
   test-owned one.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Thomas Müller <1005065+DeepDiver1975@users.noreply.github.com>

* fix(js-tests): restore $.fn.slideUp stub in fileactionsSpec

The 'cleans up after hiding' test stubbed $.fn.slideUp globally but
never called slideUpStub.restore(). When Jasmine 5 ran this test before
any test that relies on slideUp (e.g. coreSpec's menu toggle test), the
stub remained active and prevented OC.hideMenus() from hiding elements
via slideUp, causing navigation visibility assertions to fail.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Thomas Müller <1005065+DeepDiver1975@users.noreply.github.com>

---------

Signed-off-by: Thomas Müller <1005065+DeepDiver1975@users.noreply.github.com>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>