plparse/totem-pl-parser.c:1308: Warning: TotemPlParser: totem_pl_parser_save_finish: unknown parameter 'async_result' in documentation comment, should be 'result'

Use TOTEM_PL_PARSER_VIDEOSITE_SCRIPT_DIR to override the default
directory to find videosite scripts in.

Upstream is dead (last upstream change in 2016), and it doesn't
compile with lua 5.4.

And update contents for latest changes.

Both as documentation and as a placeholder for scripts in the future.

Something in the stack dropped the dependency on shared-mime-info,
breaking the is_probably_dir() detection when resolving relative URLs.

Add that missing dependency so that "foo.html" is detected as a known
filetype that's not a directory, fixing the resolution.

Working around this problem in libabigail's abidiff:
https://sourceware.org/bugzilla/show_bug.cgi?id=27267

And installing an older version to fix this one:
https://sourceware.org/bugzilla/show_bug.cgi?id=27269

Previously all XSPF playlists had the name "title".

This is to handle rss feeds with shorter subtitle than description
(or vice-versa), where the shorter subtitle occurs after the longer
description overwriting the full description with a partial
subtitle. We now always look for the longer description of
<description>, <itunes:summary> and <itunes:subtitle> for feeds.

Note that <description> contains either the full HTML (or a shortened
version if a <content:encoded> element is available, <itunes:summary> is
a plain-text <description> and <itunes:subtitle> shortened, plain-text
<description>, so <description> should always be getting selected over
those shorter versions.

This is to handle rss items with description tags with different
descriptions. We now look for the longer description of <description>
and <itunes:summary> for items. We avoid looking for item description
in <itunes:subtitle> as it could contain the full text subtitle of the
episode.

<image> tag doesn't have any node data, so the existing code was
providing image uri to podcast feed requests only from <itunes:image>
tag, if it occurs after <image> tag, returning NULL otherwise. We now
fetch image uri from <url> child tag of <image> tag.

We now look for item image uri in <itunes:image> and <image><url> tags
in addition to other tags. We don't report an error if the image url
from the tags differ. The url in the later occuring image tag replaces
the previous one.

We expect that <media:content type='...'> and <enclosure type='...'>
content-type values to be same for a podcast item. We don't report an
error if the content-type from the tags differ. The content-type in
the later occuring tag replaces the previous one.

For rss feed channels, <lastBuildDate> and <pubDate> indeed have
different meanings. Please refer [1] for more details. But podcast
providers use it in their own ways. Some use only <pubDate> (
e.g. podbean ), some only <lastBuildDate> ( e.g. anchor.fm, buzzsprout
), some use both ( e.g. podigee, soundcloud ). Podigee seems to use
<lastBuildDate> for any modifications to feed contents ( including
episode additions ), which infact should be covered by <pubDate>.

In short, we are only interested in the recent timestamp of both these
tags, when both are available.

[1] https://www.rssboard.org/rss-profile#element-channel-lastbuilddate

The order of preference for author is as follows:

1. <author> or <itunes:author>
2. <itunes:owner> -> <itunes:name>
3. <generator>

This is necessary as GNOME 3.38 now supports parental controls.

Closes: #28

This should be useful for apps which need the feed content type. This
is currently required for OPML feeds, since they need to be handled
differently by the apps than other feeds.

Don't try to convert from "utf-8" to "UTF-8" encodings, as those are one
and the same, and iconv will fail when trying that.

Encodings are defined in:
https://www.iana.org/assignments/character-sets/character-sets.xhtml

Incorrect commit message.

This reverts commit de23ed72.

Don't try to convert from "utf-8" to "UTF-8", as encodings aren't
case sensitive.

Encodings are defined in:
https://www.iana.org/assignments/character-sets/character-sets.xhtml

We shouldn't blindly return data as UTF-8 simply because the XML header
says that it is UTF-8.

Use uchardet when available to detect the encoding of XML data when the
declared encoding doesn't match the data passed.

The test expected us to pass invalid UTF-8 through, and just feed it to
the calling application.

GLib APIs actually guarantee that data passed is UTF-8 unless stated
otherwise, so make sure that no data is parsed if the data isn't UTF-8
and we can't detect what encoding it's in.

This reverts commit 91adcb2d.

Closes: #31

This change properly null terminates the CDATA values. For shorter
CDATA values, this fix is mandatory to override the token buffer text
which contains the previous token value of '<![CDATA['. CDATA value is
overwritten on top of this CDATA marker causing garbage data.

For longer CDATA values, this wasn't an issue, as the token buffer (
64K ) is already filled with NULL chars, causing this issue to not
show up.

Closes: #30

This includes needle checks for rss / atom / opml docs.

This is a partial revert of 2008 commit 3247aa1c, since node element
names do not carry whitespaces with them.

Make sure to pass the debug flag down from the parser to the videosite
helper.

Replace the quvi "videosite" script as used in tests by a mock script
that mimicks the behaviour of quvi. This allows us to test the videosite
code rather than scripts which we don't ship anyway.

libquvi has been dead upstream for a number of years, and recent lua
changes made it drop from distributions. It wasn't used in our official
distribution channel, through Flathub.

Its advertised license (GPLv2) isn't compatible with the package license
(LGPLv2.1+) so remove it.

Remove outdated msgstr in a number of translations so we don't have a
commented-out copies of the GPL blurb there.

This is useful for debugging invalid XML documents. Provides the
following debug information.

Invalid byte sequence in conversion input: byte offset 22493, byte:
'\xe2', byte context: 'ience from Valentine\xe2'

This tests for invalid byte sequences which do not fit in any charset.

It was removed in d215a84e

Rather than Bugzilla.

Contact can be made through GitLab.

We don't need to check whether links in RSS feeds are videosite links
unless the application requested that we do, by enabling recursive
parsing.

Closes: #25

Note that this changes the result of the
test_parsing_needle_carriage_return() ATOM test which now returns a
single entry.

Can be fetch from the git archives if needed.