Introduce helper functions to access the various fields of the xchan
stored in `App::$observer'. This removes direct access to the attribute
from core, with the aim of allowing further refactoring later.
We can not yet make the `App::$observer` attribute private, though, as
it is also accessed directly by some addons.
The PHP Epub Meta library has a dependency that prevents deployment on
32bit architectures. We also don't need all the functionality in that
library, so this patch replaces it with our own simplified code for
fetching the cover embedded in Epub archives.
We also expand the test suite and clean up some minor issues in the
Epubthumbnail class.
If the /owa endpoint received a request with a missing or invalid
Authorization header, it would return an error to the requester, but
without any message describing why it failes.
This patch adds a message to the error response, so that it will be a
bit easier to debug these issues in the future.
The owa spec includes a 'message' field in the error response, but makes
it optional. Any conforming implementations should accept a response
that includes the 'message' field.
MySQL does not support the INSERT...RETURNING clause, while MariaDB
does. This patch ensures that the test is not skipped on MariaDB, but
only on an actual MySQL system.
A common use case is to insert a record into a database table, but also
instantiate an object from the inserted data. This requires that we know
the value of any default or calculated columns that is filled in by the
database when the row is inserter.
This patch adds a `insert` method to pda_dbo that will insert a row, and
immediately fetch the row back from the database – including the default
and calculated values not specified by the insert itself.
This makes the configuration used align better with how it is being used
in the mdpost addon. This also reveals some issues that are less than
ideal for Markdown posts. The relevant test cases have been adjusted to
pass with the new config, but have been commented.
Only tests the delegate functionality at the moment.
This patch also includes some minor fixes to the Magic module, triggered
by the tests with empty destination URL's.
When passing a content throught the `markdown_to_bb` function
to convert any markdown in the content, any recognized URLs in the
content would be converted to BBCode links as a post processing step
after the main conversion.
After commit a1ccacb825 this did no longer
consider content within code blocks, and would thus convert them as
to BBCode links.
Example: The following content
[code]
example url: https://example.com
[/code]
Would be converted to
[code]
example url: [url=https://example.com]https://example.com[/url]
[/code]
Prior to commit a1ccacb825, code blocks
would be protected, so this would not happen.
This patch removes the post processing step for converting plain URLs to
links completely from this routine. This functionality is in any case
covered in the actual BBCode parser where it belongs.
This will have some other side effects as well, such as images and links
created using Markdown, will not be converted to [zmg] or [zrl] tags
where that would be done automatically before. If you intend to use a
[zrl] or [zmg] tag, you now need to do so explicitly.
