fix(popover): correct positioning and sizing when html zoom is applied

[KanhaiyaPandey]

🐛 Issue #30919

When CSS zoom is applied on the html element (e.g. zoom: 1.5), the popover is rendered in an incorrect position.
Additionally, size="cover" results in incorrect sizing.

---

✅ Expected Behavior

Popover should be correctly positioned and sized regardless of the document zoom level.

---

🔧 Fix

• Normalized DOMRect values and pointer coordinates based on the document zoom factor.
• Ensures consistent calculations for positioning and sizing across zoom levels.

Files updated

• core/src/components/popover/utils.ts
• md.enter.ts
• ios.enter.ts

---

🧪 Tests

• Added E2E regression test for:
  • Popover positioning
  • size="cover" behavior under html { zoom: 1.5 }
• Test file:
  • core/src/components/popover/test/zoom/popover.e2e.ts
• Firefox is skipped since CSS zoom is not supported there.

---

▶️ How to verify

cd core
PLAYWRIGHT_TEST_BASE_URL=http://localhost:3333 \
npx playwright test src/components/popover/test/zoom/popover.e2e.ts --project="Mobile Chrome"

[vercel]

@KanhaiyaPandey is attempting to deploy a commit to the Ionic Team on Vercel.

A member of the Team first needs to authorize it.

[NyaomiDEV]

Actually, Firefox now supports CSS zoom.

[thetaPC]

Thanks for working on this! This is an important fix. I've researched how other positioning libraries handle CSS zoom and I have some questions about the current approach.

Current Approach - Good Start ✅

You're correctly:

• Reading the zoom CSS property via getComputedStyle()
• Including a fallback for older browsers
• Applying the zoom factor to positioning

Questions/Gaps to Address

1. What if zoom is applied at different DOM levels?

Your implementation checks document.documentElement.zoom. But what if a developer applies zoom at a parent level instead? How does the fix handle accumulated zoom across multiple ancestors? Have you tested zoom at different levels in the ancestor chain?

2. Where are you reading the zoom from?

Are you getting zoom from documentElement, or from the popover element's own context? These could be different. Which one is correct for positioning the popover?

3. Does size="cover" work?

The issue specifically mentions size="cover" breaks with zoom. Did you verify that the sizing calculations (not just positioning) account for zoom? How does the width/height calculation change with zoom applied?

4. Are pointer coordinates handled?

If the popover uses pointer events (touch/mouse), are those coordinate adjustments also accounted for? Or only DOMRect positioning?

5. Arrow positioning

Does the popover have a separate arrow element? If so, is its positioning also adjusted for zoom?

---

Edge Cases to Test

Before marking this ready, please verify:

• Zoom at documentElement level
• Zoom at intermediate parent level
• Zoom at multiple levels (accumulated)
• Popover with size="cover" + zoom
• Arrow alignment with zoom
• Pointer-based interactions with zoom

Research Reference

I'd suggest looking at how Floating UI solved this (PR #3492) for comparison. They handle zoom differently in some key ways that might be relevant.

[caspinos]

Hi — I'm the reporter of #30919, thanks for picking this up.

I've been testing this area against main (v9.0.1) and ran into an edge case that I think is worth flagging, because the two zoom tests here wouldn't catch it.

The offscreen adjustment isn't zoom-aware

In md.enter.ts (and likewise ios.enter.ts), everything handed to calculateWindowAdjustment() is zoom-normalized — results.top/results.left, contentWidth/contentHeight, safeArea — except bodyWidth/bodyHeight, which come straight from innerWidth/innerHeight. So the clamp compares normalized coordinates against a viewport measured in a different space: it fires later than it should, and when it does fire it clamps to the wrong edge.

I think the premise behind it is the comment in utils.ts saying innerWidth/innerHeight "remain in the unscaled coordinate space". From what I can measure it's the other way round: at a 393px viewport with zoom: 1.5, innerWidth stays 393 — identical to zoom 1 — and lines up with getBoundingClientRect(), while the actual unscaled layout width is 262 (getComputedStyle(html).width, body.clientWidth). So dividing the rects by the zoom factor moves them out of innerWidth's space rather than into it.

Repro: MD mode, Pixel 5 viewport, html { zoom: 1.5 }, popover --width: 120px, trigger near the right edge of the 262px layout viewport. The popover's right edge lands at 450px on a 393px viewport — 57px offscreen. (iOS mode: 405px.) Dividing innerWidth/innerHeight by the zoom factor clamps it back on screen.

What makes this easy to miss: alignment-to-trigger and size="cover" width both still pass in that state, so a suite covering those two stays green while this is broken. That's exactly what happened in my own first attempt at this fix.

For transparency: the code path above is from this PR's head, but I confirmed the behaviour on an equivalent local implementation rather than by running this branch.

On the review feedback

While testing I ended up with working code for three of @thetaPC's points:

• Zoom at different DOM levels — using currentCSSZoom, which returns the cumulative effective zoom for the element, rather than reading zoom off documentElement (that only reports the value set on that one element, so it misses zoom on an intermediate ancestor — with nested 1.5 → 2 → 1.25, currentCSSZoom gives 3.75 where getComputedStyle().zoom gives 1.25).
• Pointer coordinates — clientX/clientY normalized for reference="event".
• Arrow positioning — arrow dimensions normalized alongside the rest.

Plus unit tests for the zoom detection and an E2E case covering the offscreen behaviour above.

@KanhaiyaPandey happy to hand any of that over if it's useful — a PR against your branch, or just the diff in a comment, whichever you prefer. Not trying to duplicate your work, I'd just like to see this fixed.

[thetaPC]

@caspinos thanks for digging into this. The offscreen clamp finding is a good catch, and the items you listed (cumulative zoom, pointer normalization for reference="event", arrow positioning, tests) are the right scope.

@KanhaiyaPandey hasn't been active here in a while, so rather than wait longer: please open a new PR with your work and we'll review it there.

Two asks for the new PR:

• Add a Co-authored-by: trailer crediting @KanhaiyaPandey, since the original approach started here. Their commit email is on the commits in this PR.
• Reference this PR and bug: Popover incorrectly rendered when css zoom is set on html element #30919 in the description.

If a PR isn't up by September 8, I'll open one on our side so this doesn't sit any longer.

We'll close this one out once a replacement is open.