Socket.IO merges unicode combining accent marks during transport

[coreh]

Hey there,

I've been hunting a bug pertaining filenames with special characters on my project, nide, and I think I've traced it back to Socket.IO.

It seems whenever I send to the server strings containing combining accent marks, such as 'COMBINING ACUTE ACCENT' (U+0301) they're somehow merged with the character before and transformed into the "normal" accented character. That is, if I call:

socket.emit('test', 'a\u0301')

The server will actually receive the string 'á' instead of the string 'a\u0301'. This is kind of hard to notice, as when printed they both look the same. The problem is that they have different lengths, and do not compare as equal.

'a\u0301' == 'á' // evaluates to false

I'm not sure if it's relevant, but I'm currently using the xhr-polling transport.

Is this behavior already known or expected? If so, how can I work around it?

Thanks in advance.

[dshaw]

What version of Node are you running against?

[coreh]

I'm running it against node v0.6.3.

[dshaw]

Could you test this against an instance of node < v0.6.2 for me?

[coreh]

Tryied it under node v0.6.1. The problem continues.

[dshaw]

Thanks.