df --local

[pajod]

filesystem remote-ness appears to be lacking tests. coreutils expanded the internal whitelist so there are expected differences to be tracked.

coreutils/src/uucore/src/lib/features/fsext.rs

Lines 412 to 417 in 67cc7a1

	#[cfg(all(unix, not(any(target_os = "aix", target_os = "redox"))))]
	fn is_remote_filesystem(dev_name: &str, fs_type: &str) -> bool {
	dev_name.find(':').is_some()
	|| (dev_name.starts_with("//") && fs_type == "smbfs" || fs_type == "cifs")
	|| dev_name == "-hosts"
	}

Does not break anything on Ubuntu 26.04 as it currently reverts to df (GNU coreutils) 9.7 due to separate bugs.

https://github.com/coreutils/gnulib/blob/71801d72c71b6ae706d06a3a9fcb9c2ab1b4e5b4/lib/filesystem-remote.c#L24-L28

[oech3]

Can we list them without looking at the referenced GPL code?

[pajod]

Can we list them without looking at the referenced GPL code?

CI can replicate a good chunk, once there is a test compares df --local (and df --all) between implementations. Well, if the test machine pretends to have all the common ones mounted (e.g. all compiled-in or available to CI kernel { cut -f2 </proc/filesystems ; find /lib/modules/$(uname -r)/kernel/fs | xargs -L1 basename | cut -d. -f1; } | sort | uniq).