Crypto - sign.update, verifier.update default parameter for strings 'utf8' not really true?

[WhatFreshHellIsThis]

• Version:v4.4.5
• Platform:Linux debian-john 3.16.0-4-586 deps: update openssl to 1.0.1j #1 Debian 3.16.7-ckt25-2 (2016-04-08) i686 GNU/Linux
• Subsystem:crypto

In the docs here:
https://nodejs.org/api/crypto.html#crypto_sign_update_data_input_encoding

It states:
"If encoding is not provided, and the data is a string, an encoding of 'utf8' is enforced."

I found if I didn't specify that parameter the calculated signature would be incorrect (not matching a standard implementation using Bouncy Castle crypto apis in Windows) if there were any unicode characters in the source string to be signed.

If I specified the parameter utf8 then it worked correctly both ways between Bouncy Castle's crypto apis in windows and Node crytpo under linux.

Same issue with verifier.update was found as well.

I suggest the docs should be updated to specify that it's (likely) binary by default if it's a string or a buffer or the sign.update and verifier.update code changed to act as documented.

Here is a snippet of code I was using for testing:
var sign = crypto.createSign('RSA-SHA256');
var testData = 'Что такое Unicode?';
sign.update(testData,'utf8');//<<===== The 'utf8' parameter should not be necessary as this is a string according to the docs, but it definitely is necessary.

[MylesBorins]

/cc @nodejs/crypto

[addaleax]

The default encoding has been changed to utf8 in Node v6 (#5522). The link to the docs you provided is for Node v6; Docs for v4 are at https://nodejs.org/dist/latest-v4.x/docs/api/.

I’m going ahead and closing this, but feel free to ask follow-up questions.

[WhatFreshHellIsThis]

Doh! I'm very sorry about that, I forgot how old the packaged Node is with Debian. Thank you for your patience.

Just FYI those node 4 docs actually say the same thing, perhaps those docs need to be updated?:
"If encoding is not provided, and the data is a string, an encoding of 'utf8' is enforced"

[addaleax]

Umm, no, you’re right, that is in the v4 docs.

The docs were changed in #5500, even before #5522… /cc @mscdex

[mscdex]

@addaleax #5500 was landed after #5522. The problem was that #5500 should not have been backported to v4.x. Or at least the references in it to 'utf8' should have been changed to 'binary' for v4.x.