This essentially reverts b042412416,
as YouTube seems to have fixed their servers.
At least I was able to query the YouTube endpoint around 150 times with
CURL_HTTP_VERSION_2TLS recently. They even advertise HTTP/3 support with
an `alt-svc` HTTP header now.
This unsets CURLOPT_HTTP_VERSION to let curl decide
on the version. This would support all curl versions and opens the
possibility for HTTP/3, but leads to inconsistent behavior depending
on the underlying curl version.
We don't set CURL_HTTP_VERSION_NONE explicitly, as it is always the curl
default and opens the path to let individual bridges override the HTTP
version where necessary.
Alternatively, setting CURL_HTTP_VERSION_2TLS explicitly would lead to consistent behavior
regardless of the curl version, but might uncover old curl bugs before the
developers enabled HTTP/2 by default.
Additionally, that requires at least PHP 7.0.7 (we require PHP 7.4
already) and curl 7.47.0 [1], released on Jan 27 2016 [2].
See also the discussion on https://github.com/RSS-Bridge/rss-bridge/pull/3249
[1] https://www.php.net/manual/curl.constants.php
[2] https://curl.se/docs/releases.html
* refactor: move function to class
* fix: use the computed bridge name as cache key
* refactor: extract method
* fix: set a feed item uid on errors
* docs
* fix: remove year from uid
* [core] Add html/convertLazyLoading($dom)
Looks for lazy-loading attributes such as 'data-src' and converts
them back to regular ones such as 'src', easier for RSS readers.
It also converts <picture> elements to plain <img> elements.
* [core] Document html/stripRecursiveHTMLSection()
Add documentation for that function (no code changes).
* [WordPressBridge] Use convertLazyLoading()
* [WordPressBridge] Unwrap image figures
<img> inside <figure> may not display on RSS readers.
This converts them back to <img>, without losing caption if present.
* [ZDNet] Convert lazy loading images
* [code] html/stripRecursiveHTMLSection: Fix typo
* fix: Call to a member function find() on bool
Happens when defaultLinkTo() is passed the empty string.
* fix: prevent exception in defaultLinkTo() when passed the empty string
* refactor
* fix: notice
* fix: Trying to get property content of non-object at bridges/PcGamerBridge.php line 36
* fix: better exception message
* fix: strpos(): Non-string needles will be interpreted as strings in the future. Use an explicit chr() call to preserve the current behavior
* fix: improve FeedExpander
Include the first libxml error in exception.
Give better error message if trying to parse the empty string.
Log all libxml errors if debug mode is enabled.
* error handling and logging tweak
* feat: improve logging and error handling
* trim absolute path from file name
* fix: suppress php errors from xml parsing
* fix: respect the error reporting level in the custom error handler
* feat: dont log error which is produced by bots
* ignore error about invalid bridge name
* upgrade bridge exception from warning to error
* remove remnants of using phps builin error handler
* move responsibility of printing php error from logger to error handler
* feat: include url in log record context
* fix: always include url in log record contect
Also ignore more non-interesting exceptions.
* more verbose httpexception
* fix
* fix
* refactor: search.js
* feat: use bridge description and short name in search
* fix bug in previous merge commit
Also reformat string from tabs to spaces
* refactor
* fix: bug in previous refactor
* chore: exclude phpcompat sniff due to bug in phpcompat
* fix: do not leak absolute paths
* refactor/fix: batch extensions checking, fix DOS issue
This fixes a future problem when code is placed under a namespace because `get_class($bridge)` will then return e.g. `RssBridge\Bridge\TwitterBridge` instead of the the current value `TwitterBridge`.
Also a bit refactoring of `Configuration.php`.
Eugene Molotov
Dag
Bocki
Mynacol
ORelio