[feather2222]

Contributor checklist

Administrative:

• I have read the README.
• I have read the CONTRIBUTING document and understand the caveats in the Pull Requests section.
• I have signed the Contributor Licence Agreement.

Commits and testing:

• My commits are rebased on the latest main branch.
• My commits are well-structured for review.
• My change has been thoroughly tested, and I am not aware of any regressions to existing features or behaviors.
• I have tested my contribution on these devices:
  • iPhone 16 Pro Simulator, iOS 26

---

Description

Fixes #6345.

Some translations of POLL_SELECT_LABEL_MULTIPLE (e.g. German) are long enough that the poll message bubble grows beyond its usual maximum width instead of the subtitle wrapping.

Root cause

CVText.measureLabel measures text with an NSTextContainer/NSLayoutManager pair and returns usedRect(for:). On affected iOS versions this measurement can fail to wrap long text, returning the unwrapped single-line width even though it exceeds the container width. Since the poll view lays out according to the measured size, the bubble stretches to the unwrapped width.

Comparing the screenshots in the issue: the English bubble is exactly at the maximum width (the poll option rows always span the full label width), while the German bubble is wider than that maximum by roughly the extra width of the German subtitle — consistent with the subtitle measurement returning an unwrapped width.

Fix

When the layout manager's measurement exceeds maxWidth, fall back to measuring with a UILabel (sizeThatFits), which wraps correctly. The previously test-only measureLabelUsingView helper is now compiled in all configurations for this purpose. Normal measurements are unaffected.

Testing

• Added CVTextTest/testMeasureLabelWrapsLongText, which measures the German poll subtitle at a narrow width and asserts the result wraps instead of exceeding maxWidth. It passes on iPhone 16 Pro Simulator (Xcode 26.2).
• Note: the measurement failure itself does not reproduce on this simulator/OS version, so the fallback path could not be exercised end-to-end; the test guards the contract (measured width never exceeds maxWidth).

[sashaweiss-signal]

Hi, thanks for your interest. It looks like this branch includes a commit from another PR; it'd be helpful if you could rebase that commit out.

[sashaweiss-signal]

Also, we believe this is fixed as of version 8.31 in c91f347, and I neglected to close the linked issue. Before reviewing, I'd like to confirm the issue still exists.

[sashaweiss-signal]

Note: the measurement failure itself does not reproduce on this simulator/OS version, so the fallback path could not be exercised end-to-end; the test guards the contract (measured width never exceeds maxWidth).

Ah. I'm going to close this; in the future, please ensure you are able to reproduce an issue before and verify your fix after making changes. We unfortunately can't accept speculative changes from external contributors, as it significantly increases our maintenance burden.

[feather2222]

Thanks for the context, and apologies for the noise — I should have verified the issue still reproduces on latest main before opening the PR, and kept the branch clean of unrelated commits. Understood on both points; I'll make sure any future PR reproduces the issue first and is verified after the change.