zoom positioning still wrong for the floating element itself with strategy: 'absolute' (known gap left by #3492)

Author: MILLERMARRUCreated Aug 12, 2026Updated Aug 12, 2026

#3492 fixed zoom positioning for the common cases (zoom on <html>/<body>, strategy: 'fixed'), but its own "Known gaps" section calls out that this specific combination is still broken:

  • zoom on the floating element itself with strategy: 'absolute'. Chrome reports an offsetParent that isn't the containing block across a zoom boundary. Broken on master too.

Filing this since I couldn't find a tracking issue for it and #3492 is already merged and closed, so I don't want this specific combination to get lost.

Repro sketch, mirroring the harness described in #3492's own PR body: put zoom on the floating element itself (not an ancestor), use strategy: 'absolute', and compare the computed position against the actual rendered pixels the way #3492's verification table did for the other cases (getBoundingClientRect() of the floating element vs. what computePosition() returned).

Per the PR author's own note this reproduces on Chrome at minimum. I haven't run the cross-engine matrix myself for this specific combination the way they did for the rest of the fix, so I can't say from my own testing whether it's Chrome-only or universal, just that it's explicitly called out as unresolved in the merged PR.