Sstrhsrtj

todoroki@todoroki-VJZ13B ~>printf "än"

                           echo "ä"

                           ä

ä

ä

udcc3udca4: u30b3u30deu30f3u30c9u304cu898bu3064u304bu308au307eu305bu3093

according to a UTF-16 decode tool, u30b3u30deu30f3u30c9u304cu898bu3064u304bu308au307eu305bu3093 is コマンドが見つかりません (= "command not found"), which is the correct Japanese output I expect.

From the printf and echo result, UTF-8 seems working correctly.

This happens in all shell outputs, such as ls (Japanese characters in filenames shows up in UTF-16 hex format)
EDIT: ls output was not utf-16, but something called "Octal Escape Sequence" (where 月 becomes 346234210)

ls in a directory which contains 3 folders named C, あいう, and 月 outputs:

todoroki@todoroki-VJZ13B ~/test> ls -l

total 12

drwxr-xr-x 3 todoroki todoroki 4096 10月  4 15:02  C/

drwxr-xr-x 2 todoroki todoroki 4096 10月 11 09:04 ''$'343201202343201204343201206'/

drwxr-xr-x 2 todoroki todoroki 4096 10月 11 09:05 ''$'346234210'/

(and this is weird because 月 of the file creation dates are shown correctly, while the directory name one isn't)

less vi nano behaves more strange; a file (a.txt, created with gedit) like below

あ

い

う

ä

will show as

in less (it complains "a.txt" may be a binary file. See it anyway?):

<E3><81><82>

<E3><81><84>

<E3><81><86>

<C3><A4>

in vi:

�~A~B

�~A~D

�~A~F

ä

and in nano:

 ^a^b

 ^a^d

 ^a^f

I don't remember what I had done, but it was correctly showing Japanese letters at least two days ago (and for more than 6 months).

What could be the problem and the way to recover from this?

I had accidentally updated my fish config file to read ~/.profile,

which included a line saying locale=C.

I changed this to locale=C_UTF8 and everything recovered.