mosh can't display emoji

[lucy]

Describe the issue

Mosh under msys2 can't display characters that need more than 2 bytes to encode in utf-16.

Steps to Reproduce the Problem

1. Connect to any host using mosh
2. Run printf '\U000102A1\n'
3. It prints �� instead of 🩹

Additional Context: Operating System, Screenshots

Everything seems to work properly if I use ssh instead of mosh, or mosh under wsl. Locale is set to en_US.UTF-8 on both sides.

OS: Windows 10 Pro version 21H2 19044.1706

[lifei]

+1. i face the same issue.

[Biswa96]

In my setup, printf '\U000102A1\n' does not print the emoji character but printf '\xf0\x9f\xa9\xb9\n' does. The same thing happens with and without mosh. See the attached screenshot.

[mintty]

1. \xf0... is U+1FA79 not U+102A1
2. MSYS2: printf '\U0001fa79'
   🩹
3. What is mosh?
4. As one of the printf commands above outputs two glyphs rather than one, is the shell a Windows program? What does type mosh say?

[lifei]

mobile-shell/mosh#234 (comment)
I do some research. I found the Cell contents is string type, may be this is problem.

[mintty]

If mosh is a Windows program (see my question 4 above), output is likely to be incompatible with UTF-8 handling.
Make a screen log (mintty -l log), produce the issue, and upload the log here.

[lifei]

well i took ten hours try to find what's wrong. Then I find that the following code does not work right in msys2.

#define _XOPEN_SOURCE
#include <stdlib.h>
#include <stdio.h>
#include <wchar.h>
#include <locale.h>
#include <stdio.h>
#include <string>

int main() {
    setlocale(LC_ALL, "zh_CN.UTF-8");
    wprintf(L"wcwidth of 0x269C is %d\n", wcwidth(0x269C));
	std::wstring in = L"📁💕😘😒🤦";
	wprintf(L"length of string in is %d\n", in.size());
	for (std::wstring::const_iterator i = in.begin(); i != in.end(); i++)
	{
    	wprintf(L"wcwidth = %d\n", wcwidth(*i));
	}
	wprintf(L"%ls", in.c_str());
    return 0;
}

Result in msys2

wcwidth of 0x269C is 1
length of string in is 10
wcwidth = -1
wcwidth = -1
wcwidth = -1
wcwidth = -1
wcwidth = -1
wcwidth = -1
wcwidth = -1
wcwidth = -1
wcwidth = -1
wcwidth = -1
📁💕😘😒🤦

Result in debian or WSL

wcwidth of 0x269C is 1
length of string in is 5
wcwidth = 2
wcwidth = 2
wcwidth = 2
wcwidth = 2
wcwidth = 2
📁💕😘😒🤦

but there is no libc in msys2.

$ ldd a.exe
        ntdll.dll => /c/WINDOWS/SYSTEM32/ntdll.dll (0x7fff75af0000)
        KERNEL32.DLL => /c/WINDOWS/System32/KERNEL32.DLL (0x7fff743a0000)
        KERNELBASE.dll => /c/WINDOWS/System32/KERNELBASE.dll (0x7fff72f10000)
        ADVAPI32.DLL => /c/WINDOWS/System32/ADVAPI32.DLL (0x7fff75980000)
        msvcrt.dll => /c/WINDOWS/System32/msvcrt.dll (0x7fff75350000)
        sechost.dll => /c/WINDOWS/System32/sechost.dll (0x7fff74470000)
        RPCRT4.dll => /c/WINDOWS/System32/RPCRT4.dll (0x7fff74f60000)
        msys-stdc++-6.dll => /usr/bin/msys-stdc++-6.dll (0x526840000)
        msys-gcc_s-seh-1.dll => /usr/bin/msys-gcc_s-seh-1.dll (0x5e8160000)
        msys-2.0.dll => /usr/bin/msys-2.0.dll (0x180040000)

I suggest that using another way to split string into cells would be bring a high compatibility.

[mintty]

So the emojis are displayed in your test program.
If you wonder about the lenghts, that's related to UTF-16 encoding but it's another issue.

[lifei]

mosh is a msys2 package which is like ssh.
mosh use wcwidth to deal with strings.
as the length is not good, mosh can not process emojis correctly.

[mintty]

Running ssh from MSYS: printf '\U0001fa79'
🩹
I think you should report this to mosh.

[lifei]

yes.
but if wcwitdh works good(someone say it was related to the version of libc), maybe mosh could be fine too.