fix ZFS command output uses wrong comma separator

[Egor-OSSRevival]

Motivation and Context

Currently, OpenZFS incorrectly uses a full stop (.) as the decimal separator in its human-readable output, even in locales that require a comma (,). This issue not only affects aesthetics but also breaks tools like sort -h, which rely on correct numerical formatting.

For example, running:

$ zfs list -o used | sort -h

produces incorrect sorting when using a locale that expects a comma separator.

This issue does not exist in illumos' ZFS, where numeric formatting respects the system locale.

This PR ensures that OpenZFS respects the system’s locale when formatting numerical output.

Description

This change ensures that OpenZFS uses the correct decimal separator based on the system locale.

The fix involves adding:

setlocale(LC_NUMERIC, "");

to lib/libzutil/zutil_nicenum.c, ensuring that numeric output is formatted according to the user's locale settings.

How Has This Been Tested?

The changes were tested on a system with different locales, including en_US.UTF-8 and de_DE.UTF-8, to verify that the decimal separator correctly follows the system locale. The output of zfs list -o used | sort -h was checked before and after applying the fix, ensuring that numeric sorting behaves as expected. Additional testing confirmed that the change does not affect systems where the default decimal separator is a period (.).

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Performance enhancement (non-breaking change which improves efficiency)
• Code cleanup (non-breaking change which makes code smaller or more readable)
• Breaking change (fix or feature that would cause existing functionality to change)
• Library ABI change (libzfs, libzfs_core, libnvpair, libuutil and libzfsbootenv)
• Documentation (a change to man pages or other documentation)

Checklist:

• My code follows the OpenZFS code style requirements.
• I have updated the documentation accordingly.
• I have read the contributing document.
• I have added tests to cover my changes.
• I have run the ZFS Test Suite with this change applied.
• All commit messages are properly formatted and contain Signed-off-by.

[amotin]

I am not really fresh on locales, but setting it per value printed does not sound right.

[Egor-OSSRevival]

Could you explain a bit more about what you mean and how I could do it better? Like, should I set the locale once somewhere else?

[amotin]

Like, should I set the locale once somewhere else?

Yea, I'd expect it to be set somewhere at the program start.

[Egor-OSSRevival]

zfs/cmd/zfs/zfs_main.c

Line 9171 in ecc44c4

(void) setlocale(LC_NUMERIC, "C");

Changing this line should fix issue too, but I don't know if it's good idea, maybe it's there for a reason. Could you tell me if it's okay?

[Egor-OSSRevival]

Could someone review it?

[amotin]

The location seems better, but having there "C" locale previously makes me think it was intentional. Have you looked into the history why it is there? I suppose somebody wanted predictable numbers in output, possibly for some scripting.

[TheJulianJES]

I imagine it's there to explicitly not change the output based on the local language, as ZFS is not translated in general.

Like, even on a German system, I wouldn't expect OpenZFS to suddenly use the (German) comma separator, as all other output is still in English. It makes no sense to only "translate" the numeric representation then. That's more confusing than just expecting the English output with a dot as the separator, as is the case with all other untranslated programs.

In your script, you could just use zfs list -o used | LC_ALL=C sort -h to have it expect "default" (English-ish) language C.

[cbinner]

Clearly, the zfs output is meant to be human readable, I would think of it as containing an implicit --human-readable flag (unset by -p).
Commonly used software like ls,du,df,sort,top,rsync etc. localises the decimal separator even if the header translation may be incomplete (like openzfs used to), which led me to open #16987.
This is the responsible change: #10878
Hardcoding an override for the user’s locale seems like an ugly hack – the suggestion of unsetting my locale to match what openzfs is doing seems to prove that.