For some reason, in NewPipeExtractor,
comments were loaded from JSON by YoutubeCommentsInfoItemExtractor as text,
sent via CommentsInfoItem#getCommentText to NewPipe,
where timestamps are converted to hyperlinks using Linkify:
https://github.com/TeamNewPipe/NewPipe/pull/2168
On the other hand, video descriptions are handled in NewPipeExtractor
by scraping the watch-page HTML.
There, timestamp links were previously mangled (and now properly parsed),
before being sent as HTML via YoutubeStreamExtractor#getDescription
to NewPipe (where HTML gets converted to Spanned).
The logic introduced in this commit is different from the above PR,
since it operates in the extractor, and mutates the HTML DOM
rather than identifying via regex.
Sometimes YouTube introduces recaptchas only on some pages. By adding an url to the ReCaptchaException the NewPipe app is able to use that url to load the page that originally caused the problem.
Also removed every instance of exception caught and rethrown with a different description: it makes no sense and it removes part of the useful stacktrace.
It is not always possible to get the url in the form "https://www.youtube.com/channel/...", so a not has been added whenever that happens to be the case (i.e. only in InfoStreamItems).
update gradle to version 5.1
fix sts issue for agegated videos
GOD DAMN FUCKING BULLSHIT
add duratin for controversal/age gated videos
bring back sts
remove ignores
fix ogg test
If ever YouTube changes thing again (or uses old urls for some unknown reason), this prevents the extractor from crashing.
As suggested here: 133cc032d9 (r283529811)
if a channel if deleted (thus it has an empty title), it is imported in NewPipe anyway, so that if it becomes undeleted in the future, it will be shown in the app.
(in the youtube subscription extractor)
Ignore subscriptions that have an empty title instead of throwing an error: the youtube subscription_manager XML file can sometimes contain those (i.e. deleted channels).
