Disable running tests in Firefox on Travis for regular PRs as it wasn't the case previously and is broken for some reason. We'll follow up in the future in #5626.
Update .travis.yml based on their latest infrastructure and new npm caching method.
Silence logging of sourceset test sources.
Closes#5616
When preselecting a new track based on user preference, make sure that the language is actually set and that the track we're testing is either a captions or a subtitles track.
Fixes#5553
The `PlayToggle` has one option `replay` which can show or hide replay icon. This can be set by passing `{replay: false}` as the default behavior replay icon is shown after video end playback.
Example of how to hide a replay icon
```js
let player = videojs('myplayer', {
controlBar: {
playToggle: {
replay: false
}
}
});
```
Fixes#4802
Sometimes the vjs-waiting class is removed prematurely after the player gets into a waiting state. This removes the graphic waiting spinner while the player is still waiting. Instead, we should make sure that the currentTime has updated before removing the spinner.
Instead of checking for blob urls in the generic updateSourceCaches method, check for blob urls inside of handleTechSourceset before updating the source cache.
Fixes#5504.
Follow-up for #5471
This makes the breakpoints option and `breakpoints()` method clearer and introduces the responsive option and `responsive()` method, which will turn on the breakpoints.
The return value of `currentBreakpoint()` was simplified to only ever return a string (empty if none).
Also, added convenience methods: `responsive()`, `getBreakpointClass()`, and `currentBreakpointClass()`.
This adds a breakpoints option. By default, this option is false meaning this is an opt-in feature.
When passing true, it will use a default set of breakpoints. Or custom breakpoints can be passed if users do not like our breakpoints (or previously-existing style decisions).
- Add breakpoints option.
- Adds some new (currently unused) classes: vjs-layout-medium, vjs-layout-large, vjs-layout-x-large, and vjs-layout-huge.
- Add updateCurrentBreakpoint and currentBreakpoint methods to the player.
- Update css/components/_adaptive.scss
- Add sandbox/responsive.html.example
Closesvideojs/video.js#4371
Like fluid mode, you can enable it with the class or by calling the fill
method. Calling fill() will turn off fluid mode and calling fluid() will
turn off fill mode.
Fluid mode takes precedence over fill mode.
Create a new createLogger module for better debugging. Each logger has its own log level and its own createLogger that will nest logs underneath them. `player.log` is also included, which logs the player id as part of the log statement. The history API also got a filter method.
For example:
```js
var log = videojs.log.createLogger('foo');
log('hello');
// > VIDEOJS: foo: hello
```
If Promise is available or if Promise is polyfilled, then we'll return a new Promise from `play()` that will get fulfilled with the value returned from the `play()` method. This means that on IE11, this promise will get resolved when we call `play()` even if play doesn't succeed. We will have a follow-on PR to polyfill this behavior and resolve or reject the promise for browsers like IE11 that don't return a promise from `play()`.
Trigger the change event on the next tick. This means that multiple changes to a track's mode will only result in a single change event on its associated TextTrackList rather than 3 events as it may be currently.
Fixes#5159
setSource is useful if you care to be fiddling with the source or doing some work depending on what source is set. However, sometimes, you don't need a setSource and want the middleware to always be selected.
Now, if setSource is missing, it will implicitly be included in the middleware chain.
Our setTimeout and requestAnimationFrame methods added dispose handlers so that they get cancelled if the component is disposed before they get a chance to run. However, we were only clearing out these dispose handlers if we cleared the timeout or the rAF manually. Instead make sure that we remove the dispose handler when it is no longer needed.
Fixes#5199.
Currently, VideoJS combines volume control with muted support, and these actions aren't actually the same. Muting/unmuting volume work independently from the volume control. For example, iOS doesn't support controlling volume programmatically but allows muting/unmuting volume.
This change will display the volume control panel and mute toggle button if the tech supports muting volume. The volume slider will continue to be hidden if the platform doesn't allow programmatically control volume. If neither muting nor control volume is supported, volume panel will not be displayed.
Fixes#4478.
This PR extends the `autoplay` to the player with a few options that should hopefully make working with browsers that disable unmuted autoplay by default easier.
The current boolean option will match current behavior and any unknown option will be treated as it does now. The new options are the string values `muted`, `play`, and `any`.
- `muted` will mute the element and call `play()` on `loadstart`,
- `play` will call `play()` on `loadstart()`, this is similar to the `autoplay` attribute
- `any` will call `play()` on `loadstart()` but if it fails it will try muting the video and calling `play()` again.
When providing the TextTrackDisplay component with track settings overrides, currently colour codes with only three digits are supported. This updates it so that 6 digit hex codes are also accepted.
the bug is that objects using the new-ish evented mixin cannot listen to the window object, preventing things like:
```
component.on(window, 'scroll', throttledListener);
```
This fixes that so anything that's a native EventTarget works.
The changes to source caching in #5156 introduced a regression where the source options were no longer available to plugins. This PR makes sure the cached source object retains any source options passed along.
When having a video-js embed with a class attribute, as part of the
changes to remove old IE support (#5041), we overwrote our addition of
the video-js class when it was missing. Instead, we want to make sure
that we don't override the class names again since they are already set
up correctly.
Fixesvideojs/http-streaming#100
Move the aria-live attribute to the control text element within buttons, rather than on the whole button, so it is not affected by the change of the title attribute, only by the change of the control text.
It seems like having aria-live on the button itself means that JAWS and NVDA announce the button both when the button text changes and when the title attribute changes. NVDA speaks the new label more times because it announces the button text as the label and the title as the description, so it says, for example, "pause button pause". JAWS doesn't appear to do this, so it doesn't repeat it as many times.
Partially addresses #5023
* We now trigger `sourceset` any time a `<source>` element is appended to a mediaEl with no source.
* `load` should always fire a `sourceset`
* `sourceset` should always be the absolute url.
There is no need to listen for user activity until a play is requested on the player and it just adds an extra timer for a player that hasn't started playing yet. Instead, just wait till the first `play` event.
Closes#5076.
SourceHandlers that use MSE have a problem: if they push a segment into a SourceBuffer and then seek close to the end, playback will stall and/or there will be a massive downswitch in quality. The general approach to fixing this that was discussed on slack was by setting the playback rate of the player to zero, buffering all that was required, and then restoring the previous playback rate. In my implementation, I've done this in the source handler (see: videojs/videojs-contrib-hls#1374).
From the video.js perspective, it should ensure that the UI reflects the buffering status and that the player API behaves like you'd expect -- that is to say, that it will fire seeking immediately after a call to currentTime, and it will fire seeked, canplay, canplaythrough, and playing when everything is buffered.
In Chrome/Firefox/Safari appending a <source> element when the media element has no source, causes what we think of as a `sourceset`. These changes make our code actual fire that event.
This was found due to the work done in #4660. Basically we reload the video element twice on every source with preload set to auto. This can potentially cause the same data to be downloaded twice.
If .load() is called, we don't know what the new source will be because of the asynchronous nature of the resource selection algorithm. However, we can know if the src attribute is set on the video element because that's what will be chosen. So, set the src property on the event object to be the src attribute or be the empty string. This matches how currentSrc works where it is an empty string when no source is set. So, when sourceset is triggered and src is "" in the event object, it means that the source is changing.
In addition, we're going to be releasing sourceset as an experimental feature so we could improve the API and make changes in minor releases.
Add setFormatTime for Video.js to allow users to change the time format. Example usage:
```js
videojs.setFormatTime((seconds, guide) => `${seconds}, ${guide}`));
```
Add resetFormatTime to reset the time format to the default.
Fixes#2931
The option for the player techCanOverridePoster is introduced in this commit. It allows techs to update the post whenever they like. isPosterFromTech_ is introduced as a private player field in order to track when a poster was set by a tech. This allows us to clear the poster whenever the tech is disposed of by the player.
Additionally, attempting to set the same poster more than once will have no effect / no changes will be made, since the poster is the same. This was done in order to stop triggering multiple posterchange events when calling player.poster(aPoster) with techCanOverridePoster set to true.
When a tech is disposed and a poster was set by it, unset the poster.
Pass a `canOverridePoster` option to techs to know whether techCanOverridePoster was set.
Fixes#4910.
Trigger a sourceset event whenever the source is set in the Html5 tech, including initial source. We override the video element's src and setAttribute methods so that we can trigger the sourceset event when people change the src with both the video element and our API methods.
The event object for sourceset will contain the src string that was provided at the time the sourceset was triggered. This is mostly important if a source is being set while a tech is changing.
A Tech has a featuresSourceset option that it can set to for sourceset handling. It can then call the helper triggerSourceset(src) to trigger the sourceset.
Middleware factories currently get run each time a source is set. Because middleware are assocated with a player, the factories should only run once per player.
This PR fixes it so that we associate a middleware instance with a middleware factory per player.
Each time a player is disposed, we will clear the cache of the middleware instances for that player.
Fixes#4677.
Use ResizeObserver when available for better and more performant resizing information, otherwise, fall back to a throttled resize event on an iframe that's the size of the player.
Allows a video.js user to disable this by setting resizeManager: false as an option since the component will not be initialized.
Add a debounce util.
This reverts #4800 (e0ed0b5) because we end up getting two playerresize events with the dimension methods now.
This will allow middleware to interact with calls to play() from the tech. This will require a method of indicating to middleware previously run that a middleware down the chain has terminated or stopped execution.
* Adds middleware mediator method that runs middleware from the player to the tech and a second time back up to the player. This category was created because play is both a setter(changes the playback state) and a getter(gets a native play promise if available). This also has the ability to tell whether a middleware has terminated before reaching the tech.
* Adds a middleware.TERMINATOR sentinel value that is available on the videojs object
* Adds play to the allowedMediators
* Adds paused to the allowedGetters
* Adds a sandbox example of a play mediator middleware
Chrome has started pausing autoplaying video elements when they are
moved in the DOM. Here we need to make sure that if the video started
autoplaying, after we move the element in the DOM we call play again.
Fixes#4720.
The core goal here is to make sure the following works in light of some middleware process that makes setting the source more async than next tick:
```js
player.src('...');
player.ready(() => player.play());
```
In fact, given this change, we should even be able to do:
```js
player.src('...');
player.play();
```
Unlike #4665, which would have clarified/changed the meaning of "ready", it remains a reflection of the tech's state and we make better use of the ability to queue things on that state and on the middleware `setSource` process.
A we retained a lot of references to DOM elements in various components. Here we clear it up. Also, make sure that we remove unused listeners as they can retain objects as well.
Update evented mixin to null out the eventBusEl_ after the component is disposed.
Add a feature for components, to tell it not to auto-initialize the evented mixin.
Re-enable the tests that were removed in #4640.
Add the ability to initialize Video.js with an element named video-js. This is because sometimes, seeing the native element is undesirable, plus, it's cool to have our own element.
Can be used just like the video embed.
IE9 is supported but only with dynamic sources as the source element can only be used inside of the video element.
We added a feature so that remote text tracks can auto-removed when a source changes. However, in 6.x we changed the source behavior to be asynchronous meaning that some text tracks were accidentally being removed when they weren't supposed to be.
For example:
```js
var player = videojs('my-player');
player.src({src: 'video.mp4', type: 'video/mp4'});
// set second arg to false so that they get auto-removed on source change
player.addRemoteTextTrack({kind: 'captions', src: 'text.vtt', srclang: 'en'}, false);
```
Now when the player loads, this captions track is actually missing because it was removed.
Instead of adding auto-removal tracks immediately to the list, wait until we've selected a source before adding them in.
Fixes#4403 and #4315.
Video.js players can accept a number of standard <video> element options (autoplay, muted, loop, etc), but not currently playsinline, which is now part of the [HTML spec](https://html.spec.whatwg.org/multipage/embedded-content.html#attr-video-playsinline). We should add it to the list of <video> attributes that can be provided to the player as options.
Prevent ClickableComponents re-adding event listeners each time enabled() is called.
* Keeps track of enabled state (this.enabled_)
* enable() doesn't do anything if the component is enabled, so the event handlers are not re-added
Fixes#4312
This adds a beforepluginsetup event as well as beforepluginsetup:$name and pluginsetup:$name events.
The drive behind this is improving the ability for people to make cross-plugin dependencies in a more robust manner.
Android Chrome now supports playbackRate properly, so removes the blacklist added in #3246 for newer Chrome versions.
Adds `browser.CHROME_VERSION` as a necessary evil.
No longer blacklists for Chrome 58+ -- this could possibly be fixed since 52, but 58 is all I've been able to test on and most users should keep Chrome up to date.
In the new middleware work, the way that new sources were loaded was refactored. We also recently made techs and components work either TitleCased or camelcased. There was one comparison that didn't do the proper check and cause the tech to be reloaded, even if the two techs were the same.
Previously timeupdate would fire before the video was playing, and the tech was not ready. This caused issues when preload was set to auto, because cuechange would fire before the video was even started for cues with a startTime of 0.
Wait for tech to be ready before watching for timeupdate
update unit tests to use TechFaker
Add a unit test to verify that we wait for Tech to be ready.
Increase timeout because IE8 can take more than 5 minutes to run.
Increase browser disconnect timeout so that browsers that succeed but don't report back to karma don't fail the build.
Make webpack more quiet.
Dispose players created in hooks tests which also increase stability.
This fixes a lot of the issues from #2746 by making the dialog inherit from our actual ModalDialog which now has tab focus trapping.
The Captions Settings dialog has some accessibility issues:
- Field labels and fields are not explicitly associated
- Keyboard focus does not move into the dialog when it is opened
- Keyboard focus is not trapped inside the dialog while it is open
- Keyboard focus does not return to the control which opened the dialog when it is closed
- The extent (top and bottom) of the dialog is not indicated to screen readers
- The dialog cannot be closed with the Esc key
- The meaning of '---' in the select fields is not clear
- The control to close the dialog is labeled "Done" rather than "Close"
- The purpose of the "Defaults" button may not be obvious, and its effect may not be apparent to screen reader users
- Focus highlighting (outline) on the Default and Done buttons is *very* hard to see
- Pressing the Done button doesn’t seem to have the same effect as pressing the Close (x) button; does it trigger the same focus movement
- This requirement to move it back to the triggering element is tricky, since clicking on that item in the CC menu dismisses the CC menu. I need to think about this a little more - either the menu should open again, or the focus should go to the main CC Menu Button
- The focus outline on the whole dialog goes too far to the left (all the way to the edge of the video window, not just to the edge of the dialog)
Fixes#2746.
Currently, the ARIA value of VolumeBar tracks the value of volume. It should instead track the position of the slider on VolumeBar, which tracks volume except when the player is muted, in which case it's 0.
Fixes#4064.
If the modal dialog was opened and the focus was preset inside the
player, move the focus to the modal dialog.
When the modal dialog is closed, move the focus back to the previously
active element.
When focus is inside the dialog, trap tab focus. This was inspired by https://github.com/gdkraus/accessible-modal-dialog and ally.js.
Fix the structure of elements in menus so that actionable elements are not children of actionable elements, as required by ARIA.
Remove unnecessary aria-labels on menus.
* Fixes#3986
* update `techOptions` to look for `TitleCase`/`camelCase` user tech options
* remove deprecated usage of Tech as Component
* add a unit test to verify that registerTech works
* change defaultTech_ to defaultTechOrder_
IE8 can't run the html5 tech and the mute toggle tests rely on a working
volume and mute functionality which tech faker does not have. Instead of
implementing it in tech faker, skip it on non-html5 environments.
If a user changed the volume to zero either via the mouse or keyboard, clicking unmute will now restore the volume back to this last position. Previously, the mute and volume values were completely not linked.
Fixes#3909.
Middleware registration now only accept a factory method which takes a player reference and returns some object that represents the middleware with the various methods on it.
Also, add a use to register a middleware for all types.
Advanced plugins introduced the concept of mixins and added two: evented and stateful.
This provides Components with the benefits of the stateful mixin
Advanced plugins introduced the concept of mixins and added two: evented and stateful.
This refactors Component to use the evented mixin, granting it event broadcast/handling capabilities.
Add middleware support. Middleware can function as go-between between the player and the tech. For example, it can modify the duration that the tech returns to the player. In addition, middleware allow for supporting custom video sources and types.
Currently, middleware can only intercept timeline methods like duration, currentTime, and setCurrentTime.
For example,
```js
videojs.use('video/foo', {
setSource(src, next) {
next(null, {
src: 'http://example.com/video.mp4',
type: 'video/mp4'
});
}
});
```
Will allow you to set a source with type `video/foo` which will play back `video.mp4`.
This makes setting the source asynchronous, which aligns it with the spec a bit more. Methods like play can still be called synchronously on the player after setting the source and the player will play once the source has loaded.
`sourceOrder` option was removed as well and it will now always use source ordering.
BREAKING CHANGE: setting the source is now asynchronous. `sourceOrder` option removed and made the default.
Switch to `keepTooltipsInside` by default and simplify DOM structure around the time tooltips and progress control.
BREAKING CHANGE: removal of `keepTooltipsInside` option.
* Switch to es3 preset for babel so that it runs last. Plugins run before presets and presets run in reverse order. Also, we ran into a weird bug in babel that causes `default` not to be quoted in some cases (https://github.com/babel/babel/issues/4799) which we've worked around here.
* Restore the es-shims for tests and the ie8 fallback script.
* Do a null-check around `Player.players`.
* use more round fractions (like 0.5 and 1) to avoid rounding issues.
Remove method chaining from videojs. This helps keep the methods consistent especially since play() now returns a Promise in some cases. Also, it can add some performance benefits.
BREAKING CHANGE: player methods no longer return a player instance when called. Fixes#3704.
Return the native Promise from `play()` if it exists. `undefined` is returned otherwise.
This comes in as part of the greater effort to remove method chaining.
BREAKING CHANGE: `play()` no longer returns the player object but instead the native Promise or nothing.
Prevent techs (and others) from being registered via registerComponent.
* `registerComponent` now throws if given an object that is not a subclass of Component (or Component itself).
* `registerComponent` now throws if given a non-empty string as its name argument.
BREAKING CHANGE: registerComponent now throws if no name or not a component is passed in.
`Player#tech` can now be called without passing an object into it. It no longer requires passing an object into it, so, current code will not break.
If nothing is passed in, a warning will be logged about knowing what you're doing. If anything is passed in, the warning is silenced.
This is both a change as well as a bug fix. We tried to have better awareness of when the underlying video element changed underneath us so we can dispose of the source handler but that broke some use cases of MSE. Given that we weren't able to fix it in a reasonable non-breaking and non-invasive solution, we're taking it out.
BREAKING CHANGE: remove the double loadstart handlers that dispose the tech/source handlers if a secondary loadstart event is heard.
Add a log levels and history api: `videojs.log.level()` and `videojs.log.history()`.
`.level()` will return the current level and you can also set it to be one of: `all`, `error`, `off`, or `warn`.
`.history()` will return a list of all things logged since videojs loaded. It can be disabled via `videojs.log.history.disable()` (and re-enabled with `enable()`) as well as cleared with `videojs.log.history.clear()`.
If the videojs embed code (a video element) is wrapped in a div with the
'data-vjs-player' attribute on it, that element will be used for the
player div and a new one will not be created. In addition, on browsers
like iOS that don't support moving the media element inside the DOM, we
will not need to clone the element and we could continue to re-use the
same video element give to us in the embed code.
This could also be extended in the future to change our embed code to a
div-only approach if we so choose.
Chrome has a bug where if you add a remote text track and try to use it programmatically, you won't get any cues displayed. So, just switch to always emulated mode.
Also, add IS_SAFARI and IS_ANY_SAFARI to the browsers.
This adds nativeAudioTracks and nativeVideoTracks tech options, this will disable hooking into the native track APIs. This would be useful when using videojs-contrib-hls on Edge.
* Refactored ChaptersButton, broke logic into several methods.
* Fixed the issues in #3447 about in some browsers tracks have an empty cues array instead of null. * Now we always subscribe to load event, and force an update. Also, track changes are handled, so chapters track can now be changed at runtime.
* Fixed the issue in #3447 about chapters menu items are not selectable. Now automatic update of the selected item based on player time works fine.
* Implemented the usage of the chapters track's label attribute as menu title, if it's present. If not, we fall back to the localized "Chapters" title.
* Refined the menu styling, so that vjs-menu-title telement won't get the hover effect, It would confuse users, because they might believe that the title item is a clickable item too.
This allows a user to register a new Player component with videojs to be used when videojs is called. If a player has been created already when trying to register a Player component, an error is thrown.
Fixes#3335 and #3016.
Tech#addRemoteTextTrack now accepts a second parameter - a boolean named manualCleanup which defaults to true preserving backwards compatibility. When that value is set to false the text track will be removed from the video element whenever a source change occurs.
Safari 10 automatically dedupes duplicate class names in an element. So,
our test was failing because we had an extra "foo" in the check. This is
an unlikely scenario that has browser variations, so, better to just
remove it.
Calling into the SWF too often is expensive. Current time and buffered don't actually change that often but it's very common to call them a couple times in the handling of a single event. Cache their return values for 100ms so the performance penalty of going through ExternalInterface is limited.
Adds `currentSource` and `currentSources` methods to the player that return the current source object, containing `currentSrc()` and `currentType()`, and all source objects that were given to the player.
Fixes#2443
enable() and disable() on clickable components is only cosmetic. "Disabled" implies the control should not be functional.
* Remove event listeners on disable() and add back on enable().
* Move adding listeners from constructor to enable
* Remove tabindex from disabled components and add disabled attribute to disabled buttons to prevent keyboard access.
HTML5 tech will return NaN instead of Infinity if playback has not started. Fires a durationupdate event once the reported duration can be believed if the duration is still Infinity, so controls can update.
Fixes#3079.
If a player is fluid and does not have a width set, and preload is set to none, the height of the player is zero. This includes where preload is forced to none by mobile Chrome as in #3606.
* If the player has the .vjs-fluid class when initialised, fluid is set to true, so adding the class behaves the same as {fluid: true} in the setup options.
* The fluid(bool) setter calls player.updateStyleEl_(). Otherwise it won't be triggered in createEl() if an aspect ratio is not also set.
* Corrects the test for a set videoWidth() in updateStyleEl_() - videoWidth() returns 0 if the width is unknown. This allows the default 16:9 to kick in rather than using 0:0.
* Disable HLS hack on Firefox for Android
* Fix canPlayType patching tests
* Add test to ensure that canPlayType is not patched in Firefox for Android
* Fix assertion message in Firefox for Android test
* Guard against exceptions in an event handler to stop them from breaking further processing of event handlers
* Added a test for try/catch behavior for exceptions originating in event handlers
- Get rid of our custom XHR shim. Export it as videojs.xhr.
- Updated XHR to be stubbed everywhere in tests to prevent errors.
- Added npm install to the review process
closes#2318closes#2594
This is important for enforcing the model that techs should
work the same for everything.
closes#2590fixes#2060
- Made techGet and techCall private functions
- Made loadTech, techName, and unloadTech private
- Cleaned up all other private method naming in the player
- Removed some unneeded comments
- Fixed a console error in dom tests from loading a track source
- Switched to non-fetching poster urls in tests to prevent errors
- Stubbed XHR for TextTrack tests to prevent log errors
- Fixed text track console errors that stubbing async didn't catch
because there's some async happening in tracks that makes it so
- XHR isn't even used until the test is complete
- Removed extra code
- Added player.dispose more places and fixed attributes test
