… just IE8. It seems there are still some rare cases where it's needed, and it's not expensive.
…8 where the onmove/onresize events don't always fire when elements are moved or resized, causing rendering elements to get out of sync with their targets. This is a big hammer: sets up an interval timer which fires 4 times a second (timing to be adjusted as real-world impact is measured) and makes each element check if it needs to be moved or resized. This is only done in IE8 standards mode as that's the only place it's needed.
…w would be rendered too tall, even though the dimensions and path were all set correctly. Moving the width/height setting to the end seems to work around it.
…ng it call update() instead. It has the same effect since init() is called from update() anyway, and this prevents double-execution of update() if it happens to be called first (thanks to Paul Irish for pointing that out).
…n that needs it
…hen init is not fired by the initial script execution nor by the element oncontentready handler. Fixes issue #23.
…not needed and was taking up a lot of processing time. Pretty decent performance gain.
…on-zero width, even if transparent, then IE adds an extra opaque white 1px border-like thing inside the real border. To get rid of this we have to increase the padding by the border width and then remove the border altogether. Closes #18.
…o each testcase, add github style, add .pie_hover selectors to make non-link hovers work in IE6
1) Removed 1px addition to clip bottom/right in IE8 as it is not needed and was resulting in one extra pixel of the image tiling to be shown 2) Make the bottom/right default clip values extend 1px beyond the box size, to prevent bug where a subpixel of the background was showing through.
…ows redistribution in more channels, e.g. public open source theme repositories for WordPress and Drupal which require all code to be under GPL.
…er than half the element's smaller dimension. Before this would result in a gradient focussize less than zero, causing rendering artifacts. Now it creates four shapes and each is clipped to show only a single quadrant of the shape; this allows each shape to shift its gradient focusposition across the halfway mark.
…ge of about:blank so that the entire background area remains reactive to mouse events. Fixes issue #47.
… or height of zero; this prevents a divide-by-zero error in the background-image position calculation and has a performance benefit as well. Fixes issue #30.
… longer uses any Blur filter, instead it now uses a radial gradient from opaque to transparent with the undocumented 'gradienttitle' fill type so it follows the rounded box shape. There is no longer any lag with initial rendering, updating size, or scrolling the page. It also gives better control over the size of the blur gradient so it more accurately matches the blur metrics of other browsers in most cases.
…yle is not "none".
…sking currentStyle whether the element is visible; fixes problem where the VML would disappear in IE6 after initial rendering, because the way we hide borders in IE6 is by setting visibility:hidden on the outer element.