Travis - run tests in Firefox

[cherniavskii]

It is possible to run karma tests in Firefox with a virtual screen.
Do not merge before all errors in FF are fixed.

Tests which should be fixed:

• Tooltip is opened when tapping on touch (Fix Tooltip tests for non-touch #5869)
• Tooltip is closed if not permanent when clicking on the map elsewhere on touch (Fix Tooltip tests for non-touch #5869)
• GridLayer number of 256px tiles loaded in synchronous animated grid @800x600px Loads 32, unloads 16 tiles zooming in 10-11 (GridLayer's test fail #5844)
• GridLayer number of 256px tiles loaded in synchronous animated grid @800x600px Loads 32, unloads 16 tiles zooming out 11-10
• GridLayer configurable tile pruning Loads map, moves forth by 512 px, keepBuffer = 0 (GridLayer's test fail #5844)
• GridLayer configurable tile pruning Loads map, moves forth and back by 512 px, keepBuffer = 0 (GridLayer's test fail #5844)

[cherniavskii]

Added also Chrome support

[mourner]

Looks like we'll only be able to turn this on once we sort out the failures on both browsers.

[cherniavskii]

@mourner yes

[cherniavskii]

Moved Chrome support to separate PR #5845

[cherniavskii]

Thanks @ghybs for fixing failing tests!

I think we are ready to merge this! 🎉

[ghybs]

I am glad you could solve some test issues by simply specifying the browser width in Travis config!

Now probably we should also improve those tests to positively specify the map container width, so that we do not rely on the browser width. For reference, these tests are:

• Marker.Drag drag in CSS scaled container drags a marker with mouse, compensating for CSS scale
• Canvas #events should not fire click when dragging the map on top of it
• Map.Drag mouse events change the center of the map

[ghybs]

Nice!

BTW please would you know what exactly the env var DISPLAY=:99.0 is for?

[cherniavskii]

improve those tests to positively specify the map container width

Fair enough, but should it block this PR from merging?

what exactly the env var DISPLAY=:99.0 is for?

Hmm, good question. It is required when using xvbf directly (see https://docs.travis-ci.com/user/gui-and-headless-browsers/).
But since I'm using xvfb-run wrapper, it seems unnecessary now.
I'll try to remove it :)

[cherniavskii]

@ghybs Yep, it works without DISPLAY env variable :) Thanks for your review! :)

[ghybs]

should it block this PR from merging?

No, this was mainly for reference for future improvement, sorry I should have made it clearer.

[ghybs]

That is awesome! Thx @cherniavskii!

[cherniavskii]

@ghybs Thanks for your help ! :)