- IMPROVED: ScrollTrigger's ignoreMobileResize is true by default now because a few versions back we switched to calculating the window height based on 100vh instead of window.innerHeight, so it really isn't necessary to force a .refresh() on mobile resize (address bar showing/hiding). This change improves performance on mobile devices because it skips the expensive refresh() when the address bar shows/hides.

- IMPROVED: removed a DisplayObject check in PixiPlugin so that it's more compatible with the upcoming release of Pixi.js version 8. See https://github.com/greensock/GSAP/issues/561

- IMPROVED: the anticipatePin feature works when scrolling in either direction (previously it worked when scrolling forward only). See https://gsap.com/community/forums/topic/39521-anticipatepin-in-reverse/

- FIXED: regression in 3.12.3 could cause a ScrollTrigger that has a numeric scrub value and a snap to prematurely jump to the final progress value when the snap finished rather than waiting for the scrub to finish. See https://gsap.com/community/forums/topic/39363-scrolltrigger-scrub-clip-path-flicker-when-scroll-back/ and https://github.com/greensock/GSAP/issues/567

- FIXED: regression in 3.12.3 could cause a tween with repeatRefresh: true AND a non-zero repeatDelay not to refresh. See https://codepen.io/GreenSock/pen/yLwLgNQ/db88d564d6308b9fcff7b65efb95febf?editors=1010

- FIXED: if you set a ScrollSmoother's effects() in a delayed fashion (after two ticks elapse after page load), it may not calculate offsets properly. See https://gsap.com/community/forums/topic/39380-scroll-smoother-effects-targets-jump-on-effectstrue/

- FIXED: regression in 3.12.3: if the playhead lands exactly on top of the yoyo position of a keyframed tween, it may render at the state before the animation started (just for 1 tick). See https://github.com/greensock/GSAP/issues/558

- FIXED: if you set invalidateOnRefresh: true on a ScrollTrigger that's attached to a .fromTo() tween, it may not render at the initial state on refresh(). See https://gsap.com/community/forums/topic/39386-svgdraw-window-resize-issues/

- FIXED: if you create a gsap.matchMedia() inside a gsap.context() and then revert() that Context, it may not clear out non-matching media queries (they may fire again when matching). See https://gsap.com/community/forums/topic/39481-pin-spacer-is-double-the-height-it-should-be-after-window-resize/

- FIXED: if the system clock changes backwards to an earlier time, it could cause animations to stop. See https://github.com/greensock/GSAP/issues/568

- FIXED: if a flexbox child is pinned by ScrollTrigger and pinSpacing is false, flex-basis is now set to the pixel height/width of the element to avoid common layout problems.

- FIXED: missing TypeScript parameter for ScrollTrigger.clearScrollMemory(). See https://github.com/greensock/GSAP/issues/571

- FIXED: if you set a default stagger value via gsap.defaults(), it could cause an error in ScrollTrigger if you create one with a numeric scrub. See https://gsap.com/community/forums/topic/39600-scrolltrigger-numeric-scrub-throws-console-errors-in-gsap-core-cannot-read-properties-of-undefined-reading-totalprogress/

- FIXED: if a motionPath tween had a start or end value that wraps around and would land precisely on an anchor point that is not at the very start or end of a path, it could render incorrectly. See https://gsap.com/community/forums/topic/39594-why-are-content-1-and-content-2-overlaped-also-what-is-the-best-way-to-add-a-timeline-to-a-main-tween-or-timeline/