Firefox currently seems extremely fickle - with `pan-y` if fires pointercancel as soon as a touch strays even a pixel or so vertically.
While `touch-action: pan-y` would be ideal (allowing users to scroll the page even when their finger started the scroll on the carousel), this prevents a swipe that isn't perfectly/only horizontal to be recognised by Firefox.
This commit allows nested structures for accordions. Also a part of
the documentation about data-children is removed because this
functionality didn't work and it's not applicable anymore.
Tests with the collapse accordion are also a bit adjusted to the new
situation.
* use a trailing slash when possible
* use https when possible
* remove a few redirected links
* consistently use `https://popper.js.org/`
* fix `iconUrl` in nuget files
* change Jekyll Windows guide to the official one
* Use `aria-selected` instead of `aria-expanded`
* Change tab.js to use `aria-selected` rather than `aria-expanded`
* Add `aria-orientation=vertical` to vertical tab list
* Remove dynamic tabs with dropdowns
* Fix non-interactive code examples
* Only set `aria-selected` on the `role="tab"` trigger - this stops `aria-selected` being incorrectly added to the `role="tabpanel"` itself (probably harmless, but nonetheless incorrect)
Use a virtual scrollbar as this is simpler than having a real one (overflow: scroll doesn't seem to work in Phantom), and disable it for the new test.
One test has also been altered to prevent erroneous fails when other inline styles are added to the body (e.g. overflow).
* Adjust margin for sticky elements on modal
Previously white space was visible to the right of sticky elements due to right padding being added to the body. This fixes#23661.
* Add unit tests for margin of sticky elements on modal
Part of an update to align the naming schemes across our components.
- Renames .popover-title to .popover-header
- Renames .popover-content to .popover-body
Refs: #22092
According to the docs, you just need to apply the `navbar-nav` to an
`<ul>` tag in a list-based navbar. In fact, `navbar-nav` seems to
override all the css properties set by `nav`.
Also, `nav-item` class should be used only for `<li>` tags, or in
conjunction with `<a>` tags to entirely avoid the list-based approach.
This commit removes the redundant `nav` class from `navbar-nav` `<ul>`
tags, moves `nav-item` from `<a>` tags to `<li>` tags and properly
adds `nav-item` class to `<li>` tags where missing.
Refs:
* https://v4-alpha.getbootstrap.com/components/navs/
* https://v4-alpha.getbootstrap.com/components/navbar/#nav
* Add carousel mouse listeners even if touch events enabled
- touch events are enabled not just on "mobile", just also on
touch-enabled desktop/laptop devices; additionally, it's possible to
pair a mouse with traditionally touch-only devices (e.g. Android
phones/tablets); currently, in these situations the carousel WON'T pause
even when using a mouse
* Restart cycle after touchend
as `mouseenter` is fired as part of the touch compatibility events, the
previous change results in carousels which cycle until the user
tapped/interacted with them. after that they stop cycling (as
`mouseleave` is not sent to the carousel after user scrolled/tapped
away).
this fix resets the cycling after `touchend` - essentially returning
to the previous behavior, where on touch the carousel essentially never
pauses, but now with the previous fix it at least pauses correctly for
mouse users on touch-enabled devices.
includes documentation for this new behavior.
* Remove aria-expanded from collapse.js target element
aria-expanded="true"/aria-expanded="false" only applies to the trigger,
not the element that is being expanded/collapsed.
* Tweak collapse.js accessibility section
...to make it clearer that the aria-expanded attribute always just goes
on the control.
* Fix collapse.js unit tests
- reword some of the text to make it clear we're checking behavior of
trigger/control
- move incorrect aria-expanded out of the <div>s and to the actual
trigger/control <a>s
- fix incorrect test assertion text output false -> true
* Only change aria-pressed if it's not an input-based radio or checkbox group
aria-pressed="true"/aria-pressed="false" is really only useful for
making on/off toggles out of, say, `<button>` elements. the attribute is
useless (and potentially confusing/conflicting) on, say, `<label>`
elements for an existing `<input type="radio">` or similar.
* Add unit test for buttons.js and radio/checkbox inputs in button groups
Fixes 2 bugs:
1. All keydowns were being prevented. Because of that the user wasn't able to navigate in the whole page using ARROW_UP/ARROW_DOWN.
2. Even when is an input or textarea the keydowns were being prevented. Because of that the user wasn't able to type any text on these elements.
When the rubberband effect causes Safari to scroll past the top of the
page, the value of scrollTop becomes negative. If the offset of the first
ScrollSpy target is 0 - essentially if the target is at the top of the
page - then ScrollSpy should not clear the active item. Conceptually, the
first item should remain active when rubberbanding past the top of the
page.
This commit fixes issue #21055 by verifying the first scrollspy target is
not at the top of the page before clearing the active nav-item.
* Use $.one() instead of $.on() since there are no $.off()s in the code.
* Remove unnecessary namespacing of listeners for the `scroll` & `load` events.
These are vanilla DOM events (not custom jQuery namespaced events)
and we're not using jQuery namespacing to manage these event listeners either (e.g. `$.off()`).
[skip validator]
Bootstrap’s .button styles can be applied to other elements, such as labels, to provide checkbox or radio style button toggling.
When the checkbox or radio state is changed, there should be triggered the change event. Currently, the change event is triggered on the Button, which is not correct. Only input fields do support the change event.
When a DOM node is passed to an HTML tooltip, the `title` node is only
moved if it is not already in the tooltip. Otherwise, `empty()` is used
instead of `detach()` before appending the `title` to avoid memory
leaks. If a DOM node is passed to a plain text tooltip, its text is
copied via jQuery `.text()`.
Replaces `.detach()` with `.empty()`, as `.detach()` is almost never
useful but instead leaks memory. The difference between `empty` and
`detach` is that the latter keeps all the attached jQuery events/data.
However, since we do not return the previous children, the user would
have to keep these themselves, thus they can `detach()` if necessary.
This is a port of https://github.com/twbs/bootstrap/pull/14552 to v4.
as role="menu" is a very specific (and strict) ARIA pattern for
desktop-like application menus, and our dropdowns are often used
as pure navigation dropdowns, this change abandons ARIA menus for
a more open-ended and light-weight approach
(see http://heydonworks.com/practical_aria_examples/#submenus and
http://www.w3.org/WAI/tutorials/menus/flyout/#improve-screen-reader-support-using-wai-aria)
note that in dropdown.js, switched to now target ``.dropdown-menu``
instead of ``role["menu"]`` - this also prevents bootstrap scripts
from "bleeding" into non-bootstrap components on the same page.
also removed the ``role=["listbox"]`` part, which appears to be
vestigial/unused (only place in bootstrap that uses that
role are carousels, and their key handling is done separately)
Overall logic for this test appears broken, possibly relating to an
older version of Bootstrap that did not require explicit
`data-toggle="button"` on single toggle buttons?
