This one is quite dumb.
`OsuGame` uses
[`loadComponentSingleFile`](https://github.com/ppy/osu/blob/15878f7f9fc7088494d3b66e98a7bc1004a1a06d/osu.Game/OsuGame.cs#L1228)
to load the `ChannelManager`. Importantly, this process does _not_ add
the component to any place in the hierarchy where it would normally be
disposed - this includes `InternalChildren`, but _also_ a lesser known
list of [currently-loading
components](https://github.com/ppy/osu-framework/blob/cfb0d7b4b673583f0cf56273e94352769aa5bc9a/osu.Framework/Graphics/Containers/CompositeDrawable.cs#L316-L323)
(those which have been sent through a `LoadComponentAsync` call).
The end result of this is that, `ChannelManager` creates the
`IChatClient` in its constructor, expecting to be able to dispose it,
but `Dispose` is never called!
And the failure case here is that `PollingChatClient` creates a
background task to continuously poll the API, unfortunately keeping a
reference to the rest of the world in the process.
Closes https://github.com/ppy/osu/issues/33455.
The fundamental misunderstanding and source of confusion in
https://github.com/ppy/osu/pull/33062 is that solo wants to show
*maximum combo*, and multiplayer wants to show *current combo*, for
their own, valid reasons. Which is spelled out explicitly in this
change.
Closes https://github.com/ppy/osu/issues/33465 probably.
This reverts the replay frame de-duplication logic to what it was before
https://github.com/ppy/osu/pull/33148#discussion_r2091549388.
I don't have good reproduction steps. I tried to write a test case for
this that isn't just "press and release a key in the same frame",
thinking that maybe there was some loophole in the osu! touch input
mapper that may produce this situation artificially, but I could not in
many configurations. So I have to assume that this just *can happen*
organically.
Reported via e-mails.
The web-side change that wasn't ported here is
https://github.com/ppy/osu-web/pull/11151. I wanted to port it at the
time but then ran into issues because this logic should *ideally* also
be applied to the beatmap set overlay leaderboards, but those are
hard-glued to `ScoreInfo`, cannot be easily weaned off it to use
`SoloScoreInfo` instead, and I did not want to make `ScoreInfo` even
more of a mess than it already is.
This time I'm just ignoring it and adding a TODO instead because I have
no confidence I will get review eyes on any refactor of the beatmap set
overlay. All I'll say that such refactors would have potentially
beneficial effects on results screens too which also (ab)use
`ScoreInfo`.
Fixes startup sounds from potentially being fetched from the wrong
source if API connection establishment takes longer than the intro
screen takes to load.
Closes https://github.com/ppy/osu/issues/22492.
As in, `https://osu.ppy.sh/multiplayer/rooms/{id}` links, clicked from
the chat overlay, now directly open in the client.
Additionally, `osu://room/{id}` can be used in the same way to open a
room from a third-party application or a browser.
Intends to close https://github.com/ppy/osu/issues/32859.
The difference between this and https://github.com/ppy/osu/pull/32942 is
that this PR takes the approach of completely moving the score sorting
behaviour to `IGameplayLeaderboardProvider` implementations.
This is going to be helpful for further work - to be precise, I am
looking to add a leaderboard position indicator in the bottom right of
multiplayer player to match stable, and having the position in the
provider will make the implementation of that *much* easier.
The rationale for this change is that the return value was mostly
useless, and at worst, misleading.
When using `LeaderboardManager`, it's assumed that a consumer will bind
to the global `Scores` list to ensure they receive updates for things
like local score changes via the internal realm subscription. If one
decides to instead use the return value of the task, it will be a static
snapshot that potentially becomes stale in the future.
I fell into this trap when refactoring the new leaderboard component
(while attempting to assert correctness that the values we are
displaying were in fact from the fetch operation we requested).
In the interest of keeping things simple, removing the return value
seems to be the best path forward.