Currently, there's a period where the API is `Offline` even though it is
about to connect (as soon as the `run` thread starts up).
This can cause any `Queue`d requests to fail if they arrive too early.
To avoid this, let's ensure the `Connecting` state is set as early as
possible.
I noticed in passing that in a very edge case scenario where the API's
`run` thread doesn't run before it is loaded into the game, something
could access it and get a guest `LocalUser` when the local user actually
has a valid login.
Put another way, the `protected HasLogin` could be `true` while
`LocalUser` is `Guest`.
I think we want to avoid this, so I've moved the initial set of the
local user earlier in the initialisation process.
If this is controversial in any way, the PR can be closed and we can
assume no one is ever going to run into this scenario (or that it
doesn't matter enough even if they did).
What I think was happening here is that the dump of the accuracy
counter's state was happening too early. The component is loaded
synchronously into the `ISerialisableDrawableContainer` before its
default position is set via the "apply defaults" `ArgonSkin` flow
- so the test needs to wait for that to take place first.
Closes https://github.com/ppy/osu/issues/25608.
Logic mostly matching stable. All operations are done on `ComboOffset`
which still makes overridden combo colours weirdly relatively dependent
on each other rather than them be an "absolute" choice, but alas...
As per stable, two consecutive new combos can use the same colour only
if they are separated by a break:
52f3f75ed7/osu!/GameModes/Edit/Modes/EditorModeCompose.cs#L4564-L4571
This control is only available once the user has changed the combo
colours from defaults; additionally, only a single new combo object
must be selected for the colour selector to show up.
As I look into re-implementing the ability to choose combo colour for an
object (also known as "colourhax") from the editor UI, I stumble upon
these wretched ternary items again and sigh a deep sigh of annoyance.
The structure is overly rigid. `TernaryItem` does nothing that
`DrawableTernaryItem` couldn't, except make it more annoying to add
specific sub-variants of `DrawableTernaryItem` that could do more
things.
Yes you could sprinkle more levels of virtuals to
`CreateDrawableButton()` or something, but after all, as Saint Exupéry
says, "perfection is finally attained not when there is no longer
anything to add, but when there is no longer anything to take away."
So I'm leaning for taking one step towards perfection.
- Closes https://github.com/ppy/osu/issues/31423.
- Regressed in https://github.com/ppy/osu/pull/30411.
Admittedly, I don't completely understand all of the pieces here,
because code quality of this placement blueprint code is ALL-CAPS
ATROCIOUS, but I believe the failure mode to be something along the
lines of:
- User activates juice stream tool, blueprint gets created in initial
state. It reads in a mouse position far outside of the playfield, and
sets internal positioning appropriately.
- When the user moves the mouse into the bounds of the playfield, some
positions update (the ones inside `UpdateTimeAndPosition()`, but the
fruit markers are for *nested* objects, and
`updateHitObjectFromPath()` is responsible for updating those...
however, it only fires if the `editablePath.PathId` changes, which it
won't here, because there is only one path vertex until the user
commits the starting point of the juice stream and it's always at
(0,0).
- Therefore the position of the starting fruit marker remains bogus
until left click, at which point the path changes and everything
returns to *relative* sanity.
The solution essentially relies on inlining the broken method and only
guarding the relevant part of processing behind the path version check
(which is actually updating the path). Everything else that can touch
positions of nesteds (like default application, and the drawable piece
updates) is allowed to happen unconditionally.
It proved to be too difficult to deal with the flow that clears user
states on stopping the watching of global presence updates. It's not
helped in the least that friends are updated via the API, so there's a
third flow to consider (and the timings therein - both server-spectator
and friends are updated concurrently).
Simplest is to separate the friends flow, though this does mean some
logic and state duplication.