I'm a Chinese speaking user. I noticed that, if the original mail splits MBCS strings into incomplete 
base64 encoded snippets, Kmail will also decode it incompletely. Say that, if we have a mail with base64 encoded subjects like this:

Subject: =?utf-8?B?V2luZG93cyA4IOeOsOW3suWPkeW4g+OAgiDor7fnjrDlnKjl?=
 =?utf-8?B?sLHojrflj5bljYfnuqfjgII=?=

We can see that "V2luZG93cyA4IOeOsOW3suWPkeW4g+OAgiDor7fnjrDlnKjlsLHojrflj5bljYfnuqfjgII=" as a whole it carry out the sentence "Windows 8 现已发布。 请现在就获取升级。", while "V2luZG93cyA4IOeOsOW3suWPkeW4g+OAgiDor7fnjrDlnKjl" carrys out "Windows 8 现已发布。 请现在?" (there is an incomplete UTF-8 character at the end), and "sLHojrflj5bljYfnuqfjgII=" carrys out "??获取升级。" (again an incomplete UTF-8 character at the begining).

Not only mails from Microsoft (lol) behave like this. I receive mails from Facebook apps, and they have this problem as well. In fact, I think it should be all right if you put two piece of MBCS string together before you convert them into QString (yes, from MBCS to UTF-16), rather than join each as a whole.

Reproducible: Always

Steps to Reproduce:
1. You're using UTF-8, GBK, SHIFT_JIS or any MBCS character encoding.
2. The original mail may not agree with some formal restrictions, but they're still ok.
Actual Results:  
The MBCS strings are decoded incompletely.

Expected Results:  
MBCS strings should be put together and decoded completely.

http://goo.gl/KieAq, This is my screenshot.