Last modified: 2013-10-22 21:08:35 UTC
Could be a bug in the test, or a regression. Happened recently. For all IE6-8: > module=jquery.localize > 9 tests of 17 passed, 8 failed.
I would like to make a snarky comment here about old IE versions, but I'll be satisfied with bumping the milestone.
This module powers some of the most visible extensions we use in wmf deployment. This is a regression that affects a large part of our browser support, and the regression was recently introduced in the last couple of weeks. This should definitely block 1.20, as a matter of fact, if we were running javascript tests in the first place, this would've blocked wmf-deployment and master as well.
(In reply to comment #2) > This should definitely block 1.20, as a matter of fact, if we were > running javascript tests in the first place, this would've blocked > wmf-deployment and master as well. Great! You've been great about supplying patches, I hope you'll fix this one this week. :)
Is this bug still open? It is reported as Known Issue on http://www.mediawiki.org/wiki/MediaWiki_1.20
Yes.
Krinkle: Do you still plan to work on this?
No. I'm currently working on other things. Unless given priority over that, this should be done by someone else, likely in Platform.
Lowering importance from blocker to normal, but I'll bring this up with MW Core to see what timeline we can address this in given it is a Grade B browser support issue: http://www.mediawiki.org/wiki/Compatibility#Browser
A quick git bisect turns up gerrit change I77593acf as the first revision where these tests fail. I see that patch changed "<html:msg blah />" to "<html:msg blah>" in a bunch of these tests. And putting those slashes back makes the tests pass in IE8 here. So the question is whether $( '<div><html:msg></div>' ) is supposed to work in jQuery, or whether it just happens to work in browsers other than IE6-8 even though that isn't guaranteed.
(In reply to comment #9) > A quick git bisect turns up gerrit change I77593acf as the first revision > where these tests fail. Trevor / Krinkle: ping, any comments? (Krinkle did the code review and Trevor wrote that patch.) > I see that patch changed "<html:msg blah />" to "<html:msg blah>" in a bunch > of these tests. And putting those slashes back makes the tests pass in IE8 > here. > So the question is whether $( '<div><html:msg></div>' ) is supposed to work > in jQuery, or whether it just happens to work in browsers other than IE6-8 > even though that isn't guaranteed.
(In reply to comment #9) > A quick git bisect turns up gerrit change I77593acf as the first revision > where these tests fail. Trevor / Krinkle: ping, any comments? (Krinkle did the code review and Trevor wrote that patch.)
Afaik this has always been somewhat unstable in IE. The tests may have been incomplete or incorrect. Anyhow, I don't know what would cause it. And doubt Trevor will either at this point. Someone'll have to dig in with a debugger and figure out where it goes wrong (last I checked it isn't a faulty unit tests but actually not working).
(In reply to comment #9) > A quick git bisect turns up gerrit change I77593acf as the first revision > where these tests fail. > > I see that patch changed "<html:msg blah />" to "<html:msg blah>" in a bunch > of these tests. And putting those slashes back makes the tests pass in IE8 > here. The commit changed all instances of <span ../></span> to <span ..></span> which was indeed a good change as the former is invalid. However it also changes some (not all) uses of <html .. /> to <html ..>. That's wrong and invalid html. > So the question is whether $( '<div><html:msg></div>' ) is supposed to work > in jQuery, or whether it just happens to work in browsers other than IE6-8 > even though that isn't guaranteed. jQuery is not supposed to fix this, garbage in, garbage out. This is handled by the browser natively. Modern browsers (HTML5 and up) tend to take this fine and just close it silently and treats it as a would-be void tag. However IE is notorious for throwing up on these. Especially in cases like <div><input></div> (no closing tag, not even self-closing) and <ul><li><a/></li></ul> (weird use of <a/> that should be <a></a>).
Change 90928 had a related patch set uploaded by Krinkle: jquery.localize: Fix incorrect use of void tag for <html:msg> https://gerrit.wikimedia.org/r/90928
Change 90928 merged by jenkins-bot: jquery.localize: Fix incorrect use of void tag for <html:msg> https://gerrit.wikimedia.org/r/90928
Is this now closeable, given the merge of Gerrit change #90928?
