Thanks for the PR! IIRC a long time ago I tried using iconv's stream interface but wasn't able to for some reason.

Are you sure this won't break anything?

Also, does the test suite pass 100% on your end?

i'd like to refactor StreamDecoder and maybe write some extra tests, but that could take a while to get me to. in the mean time, the easiest hotfix would be not to decode the stream if its unnessecary (from utf-8 to utf-8) (but it would probably still be broken for other multibyte layouts).

i haven't run tests yet.

the tests from decoder_spec.js worked out fine, i added a minimal test for the bug.

i extended the tests to include euc-jp and gb18030, omitted decoding if the charset is already utf-8 and ensured the charset found in the magic charset detection is one iconv can actually handle. (more elegant code would be possible without the magic inline charset detection though)

Beautiful. Will take a look when I have a minute.