Let's kill this pattern: nsAutoCString stringToMatch; nsresult rv = mimeConverter->DecodeMimeHeaderToUTF8( rfc2047string, charset, charsetOverride, false, stringToMatch); nsAutoString utf16String; CopyUTF8toUTF16(stringToMatch, utf16String);

After bug 998392, a lot of the strings inside JS are stored in latin1 instead of UTF16, does that play into this at all?

(In reply to Robert Kaiser (:kairo@mozilla.com) from comment #1) > After bug 998392, a lot of the strings inside JS are stored in latin1 > instead of UTF16, does that play into this at all? AUTF8String is still a nasty conversion, because it's neither UTF-16 nor Latin1.

Comment on attachment 8559275 [details] [diff] [review] Part 2: Save the needle as a UTF-16 string Review of attachment 8559275 [details] [diff] [review]: ----------------------------------------------------------------- 1) I believe that you also need to initialize mValue.utf16String in nsMsgSearchValueImpl::nsMsgSearchValueImpl(nsMsgSearchValue *aInitialValue) 2) Can't you use utf16String in nsMsgSearchValueImpl::GetStr(nsAString &aResult) to eliminate that conversion?

Comment on attachment 8559274 [details] [diff] [review] Part 1: Pass UTF-16 strings around search instead of UTF-8 Review of attachment 8559274 [details] [diff] [review]: ----------------------------------------------------------------- I have my doubts as to whether this is worth the effort, but let's try it.

https://hg.mozilla.org/comm-central/rev/1028fda75e9d https://hg.mozilla.org/comm-central/rev/4255fc99c7cf