629842 - Trunk won't load pages with "HTTP/0.9" header. Causes Linksys WRT610 status page to fail to load.

Status: VERIFIED DUPLICATE of bug 628832

(Core :: Networking: HTTP, defect)

Description

[Michael Kraft [:morac]]

I have no idea when this started since I hadn't been using my Linksys WRT610N wireless bridge until today, but if I go to the Wireless Network Status page in the nightly Minefield load, the page never loads.  There is an XML error in the error console, but that appears in Firefox 3.6.13, which loads the page fine.

I used Firebug to see what the problem was and apparently the code does two XMLHttpRequest connections to the WRT610N's internal web page.  The first one gets a response, the second one never gets a response.

If I manually go to the requested page that never gets a response (it's protected by basic authentication), it usually never loads in Minefield or Firefox 4.0b10.  It gives no error, nor will it replace the current page. Basically it's as if I never made the request.  The web console shows no response headers for the page. It says "undefined" and a time for the response if the page never loads or "NONE" if it does load (it sometimes loads if the "no-cache" header is not sent by Minefield).

The page always loads under Firefox 3.6.13, Microsoft Internet Explorer 8.0, Google Chrome 9.0 beta and Apple Safari 5.0.3, so the problem is only with the Firefox 4.0 (Gecko 2.0) loads.

I dumped the returned headers to see how they differed. The page that loads returns normal looking headers, but the page that's being ignore simply returns the following header:

HTTP/0.9 200 OK


That's it.  For whatever reason Minefield ignores the page and won't load it.  I don't know if this is by design or not, but it causes the WRT610 Wireless Network status page to fail to load.   

I tried searching for related bug in HTTP/0.9 but found none that were opened specifically against the Gecko 2.0 branch.

Since it works in all other browsers except Firefox 4, I'm filling this.

Comment 1

[Matthias Versen [:Matti]]

It would really help if you would find a regression range because there is no public available testcase.
You could start with a build that doesn't contain the fix for bug 363109

Comment 2

[Boris Zbarsky [:bzbarsky]]

Patrick just landed a patch today which has a good chance of fixing this.  

Michael, can you test with tomorrow's nightlies?

Comment 3

[Michael Kraft [:morac]]

(In reply to comment #2)
> Patrick just landed a patch today which has a good chance of fixing this.  
> 
> Michael, can you test with tomorrow's nightlies?

Looks like the patch fixed this.  I just tested in the 2011-01-29 nightly trunk load and the status page is now loading correctly.  If I had just waited a day to try this...

Any way, what bug did the patch fix?  You can dupe this bug to that one.

Comment 4

[Matthias Versen [:Matti]]

bug 628832 fixed a regression from bug 363109 (the bug that I suggested in comment #1)

Comment 5

[Boris Zbarsky [:bzbarsky]]

Michael, thanks for retesting!  Good to know this is not a problem anymore.