This commit is hopefully the final fix on Airsonic's side for #685. It
also fixes#1160, which was caused by temporary workarounds introduced
in #1080 while we were looking for a solution.
The root cause of the issue is the fact that, when we go to the next
track in an Airsonic play queue, we change the media source in the
`ended` event.
In MEJS, this translates as the following two things:
* In Airsonic's 'ended' event, we change the media source (set the `src`
attribute) and call the `load()` method, followed by the `play()`
method.
* The 'ended' event was also used internally by the MEJS player, and
one of these internal uses called the `pause()` method (presumably in
order to make sure that playback was stopped on some media renderers).
Unfortunately, the order in which these events are called depends (in
all modern browsers) on the order in which they are registered.
In our case, the first one is registered inside the `<body>` tag, but
the second one is registered with `$(document).ready(...)`. This means
that the first event handler is called before the second.
This means that, in some cases (when we're unlucky, hence the seemingly
random nature of the bug), `pause()` is called after `load()`
but before the media has finished loading.
Apparently, this causes the `AbortError: The fetching process for the
media resource was aborted by the user agent at the user's request.`
message to appear (which indicates exactly what's described in the last
paragraph), and the playback of the next song is aborted.
This is an attempt to make the Travis CI integration tests more reliable,
and changes from parallel to single forking mode.
This is a temporary measure until cucumber-jvm-parallel-plugin (which is no
longer supported) is ripped out (see #1240)
Signed-off-by: Andrew DeMaria <lostonamountain@gmail.com>
Date(Long.MAX_VALUE) is 292278994-08-17T07:12:55.807Z on Java 12, and
make Ultrasonic failed to parse.
Signed-off-by: Shen-Ta Hsieh <ibmibmibm.tw@gmail.com>
Until the podcast channel has been updated to provide it with a title, there
is no point to doing any further processing since the directory where episodes
are stored is derived from the title.
While this change is unrelated to #176, it fixes the traceback shown in that
issue.
Since the preference for the keyboard shortcuts
is set server-side, I had to shove them into a non-javascript
context, and then access them via javascript,
in a separate file.
Since I'm not a javascript expert, I'm more than open
to alternatives if this isn't the right way™ to do it.