Created attachment 653687 [details] Netvibes on Nightly The reader mode of Netvibes doesn't work anymore in current Nightlies. It displays blank page instead of feed list. To see it go to Netvibes, change its view to Reader mode and try to click on any RSS feed. I've setup a test account to demonstrate it. Go to http://www.netvibes.com/ and log in using: login: firstname.lastname@example.org password: NetvibesTest The attachments show the current view (in Nightly) and the desired view (in Opera). This is not the same bug as bug 717184.
It's a recent regression. m-c good=2012-08-04 bad=2012-08-05 http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=20fc34efd733&tochange=9453cf029b72 Bisecting needs to be done.
Why was this moved to Extension Compatibility, Jason?
Regression window(m-c) Good: http://hg.mozilla.org/mozilla-central/rev/20fc34efd733 Mozilla/5.0 (Windows NT 6.1; WOW64; rv:17.0) Gecko/17.0 Firefox/17.0 ID:20120804004627 Bad: http://hg.mozilla.org/mozilla-central/rev/0a17cde2a4b6 Mozilla/5.0 (Windows NT 6.1; WOW64; rv:17.0) Gecko/17.0 Firefox/17.0 ID:20120804111326 Pushlog: http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=20fc34efd733&tochange=0a17cde2a4b6 Regression window(m-i) Good: http://hg.mozilla.org/integration/mozilla-inbound/rev/aa100b9efcca Mozilla/5.0 (Windows NT 6.1; WOW64; rv:17.0) Gecko/17.0 Firefox/17.0 ID:20120803104127 Good: http://hg.mozilla.org/integration/mozilla-inbound/rev/87e7abe891a9 Mozilla/5.0 (Windows NT 6.1; WOW64; rv:17.0) Gecko/17.0 Firefox/17.0 ID:20120803113827 Pushlog: http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=aa100b9efcca&tochange=87e7abe891a9 Suspected: Bug 772733
Netvibes seems to use MooTools 1.2, so I'm betting it's the same as bug 781796.
You can't mark evang bugs against different sites as duplicates of each other... they need to be contacted separately, no? That said, if this problem is widespread enough then ES6 contains() might just need to change. :(
5 years ago
Mootools had an issue with Function.prototype.bind, introduced in ES5. We got through it. I'll bet we can do likewise here.
Why is bug 781796 in TE, while this bug is in JS Engine? We need to have a plan moving forward - are we going to perform outreach to all affected sites, or resolve in code?
Waldo, how exactly did Function.prototype.bind work out?
Mootools had a Function.prototype.bind implementation that was incompatible with the ES5 semantics. When we implemented ES5 semantics in bug 429507, we happened to break pages that used Mootools and expected its semantics. (I guess they overrode by testing for an existing bind function and not overriding in that case, but I'm not sure/don't remember.) The next version of Mootools changed their bind semantics to be ES5-compatible, and the few sites that were broken got fixed by the site authors, probably by updating to the fixed version of Mootools. I don't remember it being more than half a dozen sites or so that broke; we just soldiered on, roughly, and reported the breakage to the few sites that broke. The (at the time) lack of a fixed Mootools to upgrade to made it a little tricky, but we have no such problem here, as newer Mootools works fine. It probably helped somewhat that other JS engines were implementing it at the same time, so there was a semi-united front requesting sites to make fixes. Perhaps amusingly, Mootools 1.2, the last version with this contains issue, was also the last version that had non-ES5-compatible bind semantics. Versions without this contains bustage, or the bind bustage for that matter, are readily available and have been for awhile.
As far as a plan goes, I think we contact affected sites. If they move a little slower, in the short term we can just turn off the relevant methods before we do a release (but keep them on in alphas/betas/nightlies), and only put it in a release when we're happy with the site fix rate.
I am a frontend dev at Netvibes and we are aware of this issue. We are using a patched version of MooTools with an ES5 compatible Function#bind method, so this particular issue is not about this method, but the ES6 String#contains, which has a slightly different signature of the one in MooTools 1.2. String#contains is used in Element#hasClass, so it breaks the element query selector with classes. See this MooTools issue for more details https://github.com/mootools/mootools-core/issues/2402 A fix is on its way and will land soon on our production.
sounds like this will be fixed before 11/20(FF17 release), hence untracking