Feels like it's easier to understand this way. The difference of the
maximum scoring values for the entire beatmap and the max values for the
part of the beatmap that has already been played represents the act of
filling the rest of the unjudged objects with maximum results.
Closes#21920.
Weirdly enough this was semeingly fixed once before in ancient times in
3891f467a3, but then unfixed again in
566e09083f. The second change is no longer
needed since the toolbar became opaque in #9447.
- It was being glued in an ugly way that would have prevented sanely
localising it.
- Even on Linux, the filesystem (whichever one the user has chosen out
of the multitude available) still needs to support hard links for them
to have a chance of working.
This is more of a safety item. To avoid potential duplicate key in
dictionary errors (and also avoid being slightly memory-leaky), allow
`SoloStatisticsWatcher` consumers to dispose of the subscriptions they
take out.
If score submission fails, the score will not receive a correct online
ID from web, but will still be passed on to the solo statistics watcher
on the results screen. This could lead to the watcher subscribing to
changes with score ID equal to the default of -1. If this happened more
than once, that would cause a crash due to duplicate keys in the
`callbacks` dictionary.
Closes#21837.
- If the local user is restricted, then attempting to fetch their data
from the `/users` endpoint would result in an empty response.
- Even if the user was successfully fetched, their `RulesetsStatistics`
may not be populated (and instead be `null`). Curiously this was not
picked up by static analysis until the first issue was fixed.
Closes#21839.
