Fix possible crash when scaling objects in editor #32158
Merged
+32
−26
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The specific fail case here is when `s.{X,Y}` is 0, and
`s{Lower,Upper}Bound` is `Infinity`. Because IEEE math is IEEE math,
`0 * Infinity` is `NaN`.
`MathHelper.Clamp()` is written the following way:
https://github.com/ppy/osuTK/blob/af742f1afd01828efc7bc9fe77536b54aab8b419/src/osuTK/Math/MathHelper.cs#L284-L306
`Math.{Min,Max}` are both documented as reporting `NaN` when any of
their operands are `NaN`:
https://learn.microsoft.com/en-us/dotnet/api/system.math.min?view=net-8.0#system-math-min(system-single-system-single)
https://learn.microsoft.com/en-us/dotnet/api/system.math.max?view=net-8.0#system-math-max(system-single-system-single)
which means that if a `NaN` happens to sneak into the bounds, it will
start spreading outwards in an uncontrolled manner, and likely crash
things.
In contrast, the standard library provided `Math.Clamp()` is written
like so:
https://github.com/dotnet/runtime/blob/577c36cee56480dec4d4610b35605b5d5836888b/src/libraries/System.Private.CoreLib/src/System/Math.cs#L711-L729
With this implementation, if either bound is `NaN`, it will essentially
not be checked (because any and all comparisons involving `NaN` return
false). This prevents the spread of `NaN`s, all the way to positions
of hitobjects, and thus fixes the crash.
See previous commit for partial rationale. There's an argument to be made about the `NaN`-spreading semantics being desirable because at least something will loudly fail in that case, but I'm not so sure about that these days. It feels like either way if `NaN`s are produced, then things are outside of any control, and chances are the game can probably continue without crashing. And, this move reduces our dependence on osuTK, which has already been living on borrowed time for years now and is only awaiting someone brave to go excise it.
|
Interesting test failures in https://github.com/ppy/osu/runs/37991745826#r1s0; blocking pending investigation. (The fact that these failures are there now is another argument for ditching the osutk helper, as it would just silently do whatever if given invalid bounds.) |
|
Tests should be fixed with 47747ae. |
peppy
approved these changes
Mar 4, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fix possible crash when scaling objects in editor
Closes #32150.
The specific fail case here is when
s.{X,Y}is 0, ands{Lower,Upper}BoundisInfinity. Because IEEE math is IEEE math,0 * InfinityisNaN.MathHelper.Clamp()is written the following way, usingMath.{Min,Max}.Math.{Min,Max}are both documented as reportingNaNwhen any of their operands areNaN, which means that if aNaNhappens to sneak into the bounds, it will start spreading outwards in an uncontrolled manner, and likely crash things.In contrast, the standard library provided
Math.Clamp()is written like so. With this implementation, if either bound isNaN, it will essentially not be checked (because any and all comparisons involvingNaNreturn false). This prevents the spread ofNaNs, all the way to positions of hitobjects, and thus fixes the crash.Mark
MathHelper.Clamp()as banned APISee above for partial rationale.
There's an argument to be made about the
NaN-spreading semantics being desirable because at least something will loudly fail in that case, but I'm not so sure about that these days. It feels like either way ifNaNs are produced, then things are outside of any control, and chances are the game can probably continue without crashing if theNaNs are not produced. And, this move reduces our dependence on osuTK, which has already been living on borrowed time for years now and is only awaiting someone brave to go excise it.