fs: writing utf8 to fd 1 results in garbled output on windows

[mcollina]

• Version: 10.13 (also 8.x and before)
• Platform: Windows
• Subsystem: fs

The following snipped works fine on Linux and Mac OS X, but not on Windows:

const fs = require('fs')
const str = 'èè'
fs.write(1, str + '\n', 0, '', console.log)

Note that using process.stdout.write is actually working as expected.

[mcollina]

This might be related to libuv @nodejs/libuv @nodejs/platform-windows.

[bzoz]

Windows console does not speak UTF-8 (yet.. and some more details). Libuv converts UTF-8 to UCS-2 before writing to TTY:

node/deps/uv/src/win/tty.c

Line 1675 in 69cc58d

if (!MultiByteToWideChar(CP_UTF8,

[refack]

To elaborate on @bzoz

> s = new tty.WriteStream(1)
> s.write('èè\n');
èè
true

and in Windows Insider build 18282:

> const fs = require('fs')
undefined
> const str = 'èè'
undefined
> fs.write(1, str + '\n', 0, '', console.log)
èè
undefined
> null 5 èè

[mcollina]

Note that writing:

const fs = require('fs')
const str = 'èè'
fs.write(1, str + '\n', null, 'ucs2', console.log)

does not work either.

[seishun]

Is there anything Node can or should do about it? Silently converting text seems like a bad idea.

[refack]

Is there anything Node can or should do about it? Silently converting text seems like a bad idea.

IMO - Document as known limitation.

[refack]

P.S. @mcollina does the info in https://ss64.com/nt/chcp.html help (run chcp 65001 and use a Unicode enabled font)?

[mcollina]

chcp 65001 works super-smoothly. IMHO we should just document that.

const fs = require('fs')
const str = 'èè'
fs.write(1, str + '\n', null, 'utf8', console.log)

[Trott]

Fixed in 086482d