Amazon's lightning deals are not visible on Trunk. Tested on Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; en-US; rv:2.0b2pre) Gecko/20100715 Minefield/4.0b2pre, and caught via Firefox Input statements http://input.mozilla.com/search/?q=amazon+lightning&product=firefox

Regression range: http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=b7f3c1c9fb57&tochange=a5e1b195ecf7 The main culprit visible there is the TM merge.... Going to see whether that's it.

This is a site bug. The error console says: Error: assignment to undeclared variable Deal Source File: http://z-ecx.images-amazon.com/images/G/01/goldbox/client-side/gbld_js-v1._V210760631_.js Line: 368 (and in a debug build this is in fact flagged as a strict mode error). The checkin range includes this bug: Jim Blandy — Bug 514560: Forbid assignments to undeclared variables in strict mode code. r=jorendorff and the script in question in fact does "use strict;" and has this lovely code starting on line 366: if (!window.Deal) { (function(){ Deal = {}; ... )();} which is in fact something that should throw in strict mode.

Still not visible in 4.0b5pre Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:2.0b6pre) Gecko/20100905 Firefox/4.0b6pre

Still not visible in 4.0b5pre Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:2.0b6pre) Gecko/20100905 Firefox/4.0b6pre

Fwiw, I did send comments to Amazon (as a user), as well as contacting the people I know who work at Amazon. No response to the former so far, and the latter are trying to find the right people to talk to.

Boris, if your friends could inquire about the origin of this "use strict"; in case it came from some open source tool, or perhaps was based on bad advice in some public doc repository we could try to fix, that would be helpful. /be

The origin is obvious is you look at http://z-ecx.images-amazon.com/images/G/01/goldbox/client-side/gbld_js-v1._V210760631_.js -- that file is the concatenation of several third-party and Amazon JS files, and the very first one is http://www.JSON.org/json2.js which (correctly) uses strict. Then one of the Amazon files that comes after it isn't actually strict-mode compliant. This concatenation pattern is not really compatible with how "use strict" is designed.

Is anyone trying to contact amazon.com about this? They are headed into even more pain, esp. if IE9 implements strict mode. /be

I've been trying since comment 7. Sent 2 or 3 reports via their website, and my friend has been trying to get someone to respond to. It's been a black hole so far.

(In reply to comment #11) > I've been trying since comment 7. Sent 2 or 3 reports via their website, and > my friend has been trying to get someone to respond to. It's been a black hole > so far. I just pinged a contact there as well. Hopefully between the two of us we get somewhere.

My contact has filed a bug about it in their tracker, and will keep me posted. Likewise, I'll report back here.

Apparently there are several tickets on this. It's on the team's radar, but no ETA on a fix yet.

This is actually getting mentioned as a reason people are switching away from Firefox at http://lifehacker.com/5711040/browser-speed-tests-ie-9-firefox-4-beta-chromes-crankshaft-and-opera-11-beta

Kev, Can you please reach out to your contacts at Amazon. I pinged our contact as well. thanks /hja

From the webkit tracker -- https://bugs.webkit.org/show_bug.cgi?id=48377 Comment #27 From Evan Martin 2010-12-09 14:11:36 PST (-) [reply] Amazon contact says they're planning to work on it in January. Although i find myself wondering how much work is required to remove a single string that has no side effects other than breaking the site...

My guess would be a production code freeze during the traffic spike.

Getting a major e-tailer to make a change to their site during the nov-dec hypershopping season is basically impossible. If it doesn't expose them to huge liability or cost them a ton of money, it doesn't work from a risk calculus perspective, and I have a hard time faulting them for it.

(In reply to comment #19) > Getting a major e-tailer to make a change to their site during the nov-dec > hypershopping season is basically impossible. If it doesn't expose them to > huge liability or cost them a ton of money, it doesn't work from a risk > calculus perspective, and I have a hard time faulting them for it. Perhaps this was correct. They apparently fixed their code this morning, as the lightning deals now appear for me on Beta 8. http://tinypic.com/view.php?pic=2mnkbc3&s=7

(In reply to comment #4) > http://z-ecx.images-amazon.com/images/G/01/goldbox/client-side/gbld_js-v1._V210760631_.js > if (!window.Deal) { > (function(){ > Deal = {}; Site seems to be using this file now: http://z-ecx.images-amazon.com/images/G/01/goldbox/client-side/gbld_js-v39._V174650570_.js if (!window.Deal) { (function(){ window.Deal = {};

Note that Amazon's fix for this was inadequate, because the rest of the concatenated file isn't fully strict mode-safe. See bug 627531 for the second instance, which could very well not be the last one, although one hopes it is. :-\

To make things worse: The German Amazon website still uses the old version of the script that produces the error. So I tried the approach of Comment 7 and emailed them about the issue. Let's hope they come up with a "real" fix some day.