Apply CSS shorthand expansion to IE<9 only

[syranide]

Following up on #1886 and #1901, perhaps this is an edge-case that is unlikely to have a significant real-word impact, but faster is better (and a step towards isolating IE8 fixes). :)

@zpao Feel free to do as you want, but perhaps this is a solution for starting small and going from there.

Unsetting shorthand bug is reproducible on IE when document mode is 8 or less (we don't actually support IE<8, but whatever). According to all information I can find (and common sense), this only applies to IE<8 and no other browser. http://bugs.jquery.com/ticket/12385 jamietre/CsQuery#93

Warning: #2407 need to verify if this PR actually surfaces that problem even more

[syranide]

If you want, I could turn the IE8-detection into feature detection instead. Which perhaps makes more sense as per the discussion in the other PR. EDIT: But, React does use docoument.documentMode in a lot of places for the event system.

[sebmarkbage]

Yes, feature detection, please. Two wrongs doesn't make a right.

[syranide]

@sebmarkbage Well is (safe) UA-detection for a nonsensical error in a legacy browser really worth feature testing? I feel it makes a temporary workaround a lot more permanent when there is no one who knows if it's actually necessary, or when it can be safely removed (but perhaps we don't want to "find out in production" either as I assume no one definitively knows if other browsers are affected...).

We use document.documentMode in a bunch of other places when we're keenly aware that only old IE is a culprit.

But say (repeat) the word and I'll do it :)

[syranide]

@sebmarkbage I changed it to use feature detection instead.

[syranide]

Somewhat related and important: #2407

[sophiebits]

Why does this throw? All the info online I can find indicate that it silently does the wrong thing.

[syranide]

>>document.createElement('div').style.font = '';
  "Invalid argument." (error)
>>document.createElement('div').style.fontSize = '';
  ""

[syranide]

I'm gonna go with blocked until #2407 is resolved, at first glance it seems that #2407 should be fixed by the current lack of a feature test (but apparently isn't for some reason), and adding this feature test would make it more prevalent.

[sophiebits]

(That was me merging, GitHub seemed a little confused.)

[zpao]

Does this now hold the record for "oldest PR merged"?

[sophiebits]

I think so.

[sophiebits]

I've got my eye on #1510 though…