[XHR] Fix a XHR-stream test script.

LayoutTests/http/tests/xmlhttprequest/resources/slow-failure.cgi

Index: LayoutTests/http/tests/xmlhttprequest/resources/slow-failure.cgi
diff --git a/LayoutTests/http/tests/xmlhttprequest/resources/slow-failure.cgi b/LayoutTests/http/tests/xmlhttprequest/resources/slow-failure.cgi
index 8b2b92f6371bee97be4903a26e73b2c5423b7ead..77da15a89b93808ada59024bcbdae2e72e365d4b 100755
--- a/LayoutTests/http/tests/xmlhttprequest/resources/slow-failure.cgi
+++ b/LayoutTests/http/tests/xmlhttprequest/resources/slow-failure.cgi
@@ -7,8 +7,10 @@ $| = 1;
 autoflush STDOUT 1;
 

[hiroshige, 2015/06/18 08:49:28]

Please add comments describing expected behavior ("the XHR instance gets LOADING
and then get an error"), and how this expected behavior is achieved by this
script, somewhere in the script.

[yhirano, 2015/06/18 11:18:07]

Done.

 print "Content-Type: text/html; charset=us-ascii\n";
+print "X-Content-Type-Options: nosniff\n";

[hiroshige, 2015/06/18 08:49:28]

How "X-Content-Type-Options: nosniff" header works in the context of this CL?
I think Line 13 is to make loading start by outputting something not causing
errors, but not sure about this nosniff line.

[yhirano, 2015/06/18 11:18:07]

I add this option just in case, because Chrome may buffer bytes to sniff a mime
type without this option though I do not completely understand the mechanism.
See https://code.google.com/p/chromium/issues/detail?id=2016 for example.

 print "Transfer-encoding: chunked\n\n";
 
+print "5\nhello\n";
 sleep 2;
 
 print "hoge\nfuga\n";