Working around this problem would require us to skip setting whitespace-only
values except when they're valid which would be very fragile. Another option
would be to set the value and see if it succeeded and then react to that.
We've tried something like that in the past to be able to overwrite !important
styles (see 24e5879) but it broke the CSS cascade (see
https://bugs.jquery.com/ticket/14836#comment:5) and was triggering
MutationObserver callbacks too often so it was reverted in PR gh-1532.
Ref gh-3204
Ref gh-1532
This commits backports some changes done in the patch to the then-existing
compat branch that removed support for old browsers and added some support
comments.
Refs 90d7cc1d8b
Support comments that were lacking the final IE/Edge version that exhibits
the bug were checked & updated. Links to the Chromium bug tracker were updated.
Code in tests related to unsupported browsers (like Android 2.3 in non-basic
tests) has been removed.
Fixes gh-2868
Closes gh-2949
- Ignore certain tests that obviously are not supported
- Beefed up the sortOrder, uniqueSort, isXMLDoc, and attr functions
Fixes gh-1742
Fixes gh-2048
Close gh-2703
Unit test changes some uses of .show() and .hide() to .css( "display", ... ),
there was already an implicit assumption in several of the existing tests.
Fixes gh-2193
Close gh-2648
Support comments that mentioned only Safari < 7 were checked & updated
to account for bugs existing in newer versions as well; Safari 6 support
test results were removed.
Refs gh-2482
Android 2.3 chokes on unquoted reserved words being used as property names
which was making Deferred tests not run.
Acknowledge the sad fact that Android 2.3 is not ES5-compliant browser
and enable the "es3" option in JSHint config.
Fixes gh-2478
Closes gh-2481
This prevents jQuery from caching a prefixed property name if provided
directly by the user, e.g. the following code:
elem.css( "msTransform", "translate(5px, 2px)" );
should not prevent one from from later setting the transition directly:
elem.css( "transform", "translate(5px, 2px)" );
on a browser not understanding the unprefixed version which is the case
for Safari 8 & transform.
Fixes gh-2015
Closes gh-2298
- Reverts behavior from 10399dd, which we never released.
BR and inline elements are considered visible.
- The possibility of dropping .offsetWidth and .offsetHeight
was debunked by this perf:
http://jsperf.com/visible-hidden-and-getclientrects
Fixes gh-2227
Close gh-2281
IE9-10 throws on elements created in popups (see #14150), FF meanwhile throws
on frame elements through "defaultView.getComputedStyle" (see #15098)
Use "defaultView" if in the popup which would fix IE issue,
use "window.getComputedStyle" which would fix FF issue.
And everybody wins, except performance, but who cares right?
Fixes#15098
Closes gh-1583
It's not easy to find a CSS property that is implemented even in
ancient browsers and that returns a consistent result among browsers
when passed through .css(). color didn't work since Firefox normalizes
it an empty value to rgb(0, 0, 0).
The workaround to be able to change !important styles broke the browser
keeping the old CSS value if the new one was rejected. Patching it would
involve a significant perf hit (~33%) so the initial patch needs to be
reverted instead.
Tests by m_gol & gibson042.
Fixes#14836
Closes gh-1532
