Jesse Ruderman Reporter
	Description • 17 years ago

Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.9a7pre) Gecko/2007072101 Minefield/3.0a7pre

Security-sensitive because this could allow sites to get around popup blocking.

	Jesse Ruderman Reporter
	Comment 1 • 17 years ago

There's a slight difference in behavior between tabs and windows.  If you have Firefox set to open popups as windows, you'll get windows saying "x FAIL", but if you have it set to open them as tabs, you'll get blank tabs.

	Boris Zbarsky [:bzbarsky] Assignee
	Comment 2 • 17 years ago

Hmm.  This is fun.  Basically, we can't handle the load in the docshell that we started it in, but still try to do the load (this sounds like a bug to me, one which I should look into).  So we try to look for another place to handle the load, and the Firefox UI helpfully opens up a new window or tab to do it in.  The js stack to the tab opening is:

0 [native frame]
1 addTab(aURI = "about:blank", aReferrerURI = null, aCharset = null, aPostData = null, aOwner = [object XULElement @ 0x92632d0 (native @ 0x92632a0)], aAllowThirdPartyFixup = false, attrChanged = [function]) ["chrome://global/content/bindings/tabbrowser.xml":1224]
2 loadOneTab(aURI = "about:blank", aReferrerURI = null, aCharset = null, aPostData = null, aLoadInBackground = false, aAllowThirdPartyFixup = false) ["chrome://global/content/bindings/tabbrowser.xml":1078]
3 anonymous(aURI = [xpconnect wrapped nsIURI @ 0xab882748 (native @ 0x8818df4)], aOpener = null, aWhere = 3, aContext = 1) ["chrome://browser/content/browser.js":3476]
4 [native frame]
5 handURIToExistingBrowser(uri = [xpconnect wrapped nsIURI @ 0x8d392d0 (native @ 0x8818df4)], location = 0, cmdLine = null) ["file:///home/bzbarsky/mozilla/vanilla/obj-firefox/dist/bin/components/nsBrowserContentHandler.js":588]
6 bch_handleContent(contentType = "text/html", context = [xpconnect wrapped nsIInterfaceRequestor @ 0x8ccc358 (native @ 0xab8eb0e4)], request = [xpconnect wrapped (nsISupports, nsIRequest, nsIChannel) @ 0x889e340 (native @ 0xaba22720)]) ["file:///home/bzbarsky/mozilla/vanilla/obj-firefox/dist/bin/components/nsBrowserContentHandler.js":531]

	Boris Zbarsky [:bzbarsky] Assignee
	Comment 3 • 17 years ago

The loadgroup was getting canceled after we'd already started up the string channel... This makes sure we get canceled properly in that case.

Jesse, is this trunk-only?  I would assume it is...

	Boris Zbarsky [:bzbarsky] Assignee
	Comment 4 • 17 years ago

On branch you'd want to test without the 5ms timeout, probably, since javascript: is sync there.

	Jesse Ruderman Reporter
	Comment 5 • 17 years ago

I can't reproduce in Firefox 2.0.0.5.  I tried a timeout of 5ms, a timeout of 0ms, and no timeout.

	Christian :Biesinger (don't email me, ping me on IRC)
	Comment 6 • 17 years ago

Comment on attachment 273324 [details] [diff] [review]
Proposed patch

I wouldn't bother with remembering listener->OnStopRequest()s return value in OnStopRequest, errors from here are ignored anyway

	Boris Zbarsky [:bzbarsky] Assignee
	Comment 7 • 17 years ago

Fixed.