Incorrect CurrentCulture derived from OS in .NET 8 runtime

[pvlst]

I'm struggling with incorrect CurrentCulture setting coming from the operating system.

• the application is built using the SDK image and runs in the aspnet image
• used images are 8.0-bookworm-slim (Debian 12, amd64)
• deployed on K8S cluster

When preparing the SDK and aspnet images, the required locales are installed and then set via env variables:

ENV LANG "cs_CZ.UTF-8"
ENV LC_MESSAGES "en_US.UTF-8"

This should work so that all the localized values are in the cs_CZ locale and only messages (exceptions, dotnet tools messages etc.) are in the en_US locale. But it doesn't work as expected. Everything is in en_US locale format (dates, numbers, messages, ...). It's as if the cs_CZ is not set at all.

When I type the locale command into the system, it shows:

LANG=cs_CZ.UTF-8
LANGUAGE=
LC_CTYPE="cs_CZ.UTF-8"
LC_NUMERIC="cs_CZ.UTF-8"
LC_TIME="cs_CZ.UTF-8"
LC_COLLATE="cs_CZ.UTF-8"
LC_MONETARY="cs_CZ.UTF-8"
LC_MESSAGES=en_US.UTF-8
LC_PAPER="cs_CZ.UTF-8"
LC_NAME="cs_CZ.UTF-8"
LC_ADDRESS="cs_CZ.UTF-8"
LC_TELEPHONE="cs_CZ.UTF-8"
LC_MEASUREMENT="cs_CZ.UTF-8"
LC_IDENTIFICATION="cs_CZ.UTF-8"
LC_ALL=

Which seems fine. However, .NET's CurrentCulture and CurrentUICulture both return en-US. The only way to change this is to set LC_ALL to cs_CZ, but that also changes the language of messages, and of course not only in the .NET application, but also in the tools etc., which I don't want.

Why does the .NET 8 not use the cs_CZ locale for dates, numbers or anything really?
Am I doing something wrong or is this a bug?
How is the CurrentCulture determined?
Can you point me to some technical documentation or even source code that explains how it's determined from the OS settings?

Thanks in advance.

[MichalPetryka]

How is the CurrentCulture determined?
Can you point me to some technical documentation or even source code that explains how it's determined from the OS settings?

I assume it's the same as here:

runtime/src/libraries/Common/src/System/Text/EncodingHelper.Unix.cs

Lines 21 to 29 in 35e23f7

	/// <summary>Environment variables that should be checked, in order, for locale.</summary>
	/// <remarks>
	/// One of these environment variables should contain a string of a form consistent with
	/// the X/Open Portability Guide syntax:
	/// language[territory][.charset][@modifier]
	/// We're interested in the charset, as it specifies the encoding used
	/// for the console.
	/// </remarks>
	private static readonly string[] s_localeEnvVars = { "LC_ALL", "LC_MESSAGES", "LANG" }; // this ordering codifies the lookup rules prescribed by POSIX

But I can't find the exact place that does it.

Why does the .NET 8 not use the cs_CZ locale for dates, numbers or anything really?

AFAIR .NET only uses a single culture at once, it has no way to use different cultures for different things without you specifying it manually.
Additionally, it's often preferred to use InvariantCulture everywhere but on displaying data to user to avoid behavioural differences between systems.

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/area-system-globalization
See info in area-owners.md if you want to be subscribed.

[pvlst]

I assume it's the same as here:

runtime/src/libraries/Common/src/System/Text/EncodingHelper.Unix.cs

Lines 21 to 29 in 35e23f7

Thanks, I'll have a look.

Why does the .NET 8 not use the cs_CZ locale for dates, numbers or anything really?

AFAIR .NET only uses a single culture at once, it has no way to use different cultures for different things without you specifying it manually. Additionally, it's often preferred to use InvariantCulture everywhere but on displaying data to user to avoid behavioural differences between systems.

Yes, I use InvariantCulture everywhere except for the human readable output. For example, the application generates many different localized reports that need to have numbers, dates etc. in the cs_CZ format.

The en_US messages and cs_CZ values settings does work correctly in my Windows development environment, and I don't set this explicitly anywhere. So I would hope that it can work the same on Linux environment.

[Suchiman]

AFAIR .NET only uses a single culture at once, it has no way to use different cultures for different things without you specifying it manually.

Not multiple cultures but a custom / modified / forked culture with user overrides like on windows could work. E.g. on Windows you can go to your regional settings and adjust things to your liking


which then reflect in the CurrentCulture. You can explicitly opt out of that by specifying new CultureInfo("de-DE", useUserOverride: false) for example. On non windows platforms, .NET could do the same thing, by starting with a base culture as specified by LANG=cs_CZ.UTF-8 and then load overrides to specific settings.