GitLab

This has no impact in real life since the table is temporary, but it
will cause future migration tests to fail if we create a new
FileRelationTable backup table

All the information we need is available in the Media table.

We're searching for a media, it has to have a main file associated with
it, so we don't have to force a join with the file table in case we
don't sort by a File related field

This was just benching the construction of the query object...

We need the custom logger before any initialization step runs, otherwise
those logs will be lost.
In order to simplify the locking, it seems reasonable to mandate that the
logger won't be updated afterward.

Otherwise some logs could be missing, especially the
devicelister/fsfactories ones.

Because vlc-android and the medialibrary now share the same libvlc
instance, dialogs spawned in libvlc while opening a media to generate a
thumbnail will be shown to user. This should not happen, and with the
no-interact option it won't.

amrec is a synonym for MPEG TS used by certain camera systems.

Which allows application to signal that a media was played in order to
increase its playcount and position in history without providing a
playback position.
Fix #434

From the users' perspective, a media can be considered as removed from
the database as long as it's not part of any listing.
This simplifies the implementation for the clients, and matches the
initial intent of not providing explicit callbacks for external media.

Fix videolan/vlc#26854

Fix #435

refs #435

refs #435

We'll soon have a 2nd similar queue so we need to clarify what kind of
convertion that is

It's never used.

In order to update the duration when a media gets removed, similar to
what was done for the number of tracks for #432

This allows use to react to convertion to/from external media
Fix #432

Instead of manually updating when assigning a track to the genre.
This allows us to update the number of tracks automatically when a media
gets converted to external, mainly following a removal of the associated
entrypoint.
Refs #432

Fix #431

This only ever caught false positive

This reverts commit 433593c5.
This wasn't enough to fix the problem, as a refresh would:
- Invoke flush, setting the flushed flag to true
- Invoke restore, immediatly setting the flag back to false

This saves a bit of duplicated code in each template instantiation

This saves about 20KB in the resulting binary
refs #430

Now that we don't have to manually time the request, we can just return
the result directly

This prevents this common code to be generated for each variant, saving
around 10KB in the resulting binary
refs #430