Fix Native Digit Reading #58598
Fix Native Digit Reading #58598
Conversation
|
Tagging subscribers to this area: @tarekgh, @safern Issue DetailsWhen creating any culture, we read the culture data from the underlying OS library (either NLS or ICU). Part of the data we read is the culture native digits. We should get ten digits string (e.g. It is hard to add a test for such fix as it will need to change the machine settings. But, I have tested manually and ensured it address the issue when having wrong settings.
|
src/libraries/System.Private.CoreLib/src/System/Globalization/CultureData.cs
Outdated
Show resolved
Hide resolved
src/libraries/System.Private.CoreLib/src/System/Globalization/CultureData.cs
Show resolved
Hide resolved
| @@ -2138,29 +2138,33 @@ private string[] GetNativeDigits() | |||
|
|
|||
| // if digits.Length < NumberFormatInfo.s_asciiDigits.Length means the native digits setting is messed up in the host machine. | |||
| // Instead of throwing IndexOutOfRangeException that will be hard to diagnose after the fact, we'll fall back to use the ASCII digits instead. | |||
| if (digits.Length >= NumberFormatInfo.s_asciiDigits.Length) | |||
| if (digits.Length < NumberFormatInfo.s_asciiDigits.Length) | |||
There was a problem hiding this comment.
I think I'm missing something obvious -- why is this < rather than !=
There was a problem hiding this comment.
digits can include null termination character at the end.
ghost
locked as resolved and limited conversation to collaborators
When creating any culture, we read the culture data from the underlying OS library (either NLS or ICU). Part of the data we read is the culture native digits. We should get ten digits string (e.g.
0123456789). Sometime the settings in the host machine get messed up and can return digits strings less than ten digits and that can cause .NET to throwIndexOutOfRangeException. Such failure is hard to diagnose after the fact especially when the collected dumps don't include the machine settings. This issue gets reported every while and recently we get it again from some customer. The change here is only read the returned digits data if it is ten digits. Otherwise, we'll fallback to ASCII digitsIt is hard to add a test for such fix as it will need to change the machine settings. But, I have tested manually and ensured it address the issue when having wrong settings.