It don't work. E/GeckoConsole( 1004): [JavaScript Error: "viewport is null" {file: "chrome://global/content/BrowserElementChild.js" line: 867}] which is _recvViewportChange: function(data) { let viewport = data.json; and the data originates from TabChild::RecvUpdateFrame(const nsIntRect& aDisplayPort, [snip] { nsCString data; data += nsPrintfCString("{ \"x\" : %d", aScrollOffset.x); data += nsPrintfCString(", \"y\" : %d", aScrollOffset.y); [snip] return RecvAsyncMessage(NS_LITERAL_STRING("Viewport:Change"), NS_ConvertUTF8toUTF16(data)); (Yes, this is terrible.) So it looks like the mm API changed, no longer |.json|, or our manually-built string isn't a valid structured clone.

Confirmed that https://hg.mozilla.org/integration/mozilla-inbound/rev/b83188166929 works.

I did mention at some point that I may or may not break that method - but since there are no tests for it, had to guess. Apparently I did break it. Someone who knows how to test the functionality needs to fix it (and add tests!).

Comment on attachment 648286 [details] [diff] [review] Fix up shim code supporting hack to deliver mm message from C++ Uh. sorry! And thanks for fixing this. Any chance to get some kind of test for this?

This is a bustage fix, blocker. https://hg.mozilla.org/integration/mozilla-inbound/rev/be31a4c795f7

(In reply to Olli Pettay [:smaug] from comment #4) > Comment on attachment 648286 [details] [diff] [review] > Fix up shim code supporting hack to deliver mm message from C++ > > Any chance to get some kind of test for this? Yes. The situation is complicated, sadly.