Bug 16987 – XMLHTTPRequest.responseText (text/html) without charset label does not inherit the referring page's charset

• Status: RESOLVED
• Resolution: FIXED
• Component: Page Loading

[Jungshik Shin]

* How to reproduce
  1. Go to http://i18nl10n.com/webkit/xhrtest.html 

* Expected (IE6, Firefox and Opera all do this)
  Both columns should be identical with two lines of the first 5 letters of Cyrillic alphabet

* Actual
  The 2nd line in the right column (text/html) has 'gibberish' rather than the first 5 letters of Cyrillic alphabet. 



The left column ( http://i18nl10n.com/webkit/xhrtest1.html with meta for UTF-8) loads xhrtestdata1.txt (encoded in UTF-8 and emitted with HTTP header : Content-Type: text/plain) while the right column (http://i18nl10n.com/webkit/xhrtest2.html : with meta for UTF-8) loads xhrtestdata2.html (encoded in UTF-8 without meta and emitted with HTTP header : 'Content-Type: text/html')

Neither xhrtestdata1.txt nor xhrtestdata2.html has 'UTF-8 BOM'.

[Jungshik Shin]

I forgot to mention that the gibberish is the result of interpreting the 1st 5 Cyrillic letters in UTF-8 as ISO-8859-1 (the default encoding). 

There can be other variations of these two test cases.

[Jungshik Shin]

sorry for bug spam. 
(In reply to comment #1)
> I forgot to mention that the gibberish is the result of interpreting the 1st 5
> Cyrillic letters in UTF-8 as ISO-8859-1 (the default encoding). 

s/ISO-8859-1/Windows-1252/   

[Eric Seidel (no email)]

This fails for me on 3.0.4 (mac), however it works on TOT.  I think this bug was fixed already, not sure where.

[Eric Seidel (no email)]

Fails in the latest Windows beta.

This is the file which would have changed:
http://trac.webkit.org/projects/webkit/log/trunk/WebCore/xml/XMLHttpRequest.cpp

A quick scan didn't find the change.  However, I think we can still close this as "Fixed".

[Jungshik Shin]

It's 'fixed' in http://trac.webkit.org/projects/webkit/browser/trunk/WebCore/xml/XMLHttpRequest.cpp?rev=28934

However, I'm not sure that change is the 'right' thing to do.  I'll make another test case and see what happens. 

[Jungshik Shin]

I was wrong to think that 'text/html'(without charset specified anywhere) obtained through XHR inherits charset from the referring document in FF and IE. ap's change makes webkit compatible with FF and IE. 

[Jungshik Shin]

(In reply to comment #6)

> ap's change makes webkit compatible with FF and IE. 

Which assumes UTF-8 when charset is not specified.