Skip to content

Resize the layout for the on-screen keyboard in Chrome on Android - #2038

Closed
joepio wants to merge 1 commit into
developfrom
claude/mobile-meeting-chat-input-btc3g1
Closed

joepio wants to merge 1 commit into
developfrom
claude/mobile-meeting-chat-input-btc3g1

Conversation

@joepio

@joepio joepio commented Oct 3, 2026

Copy link
Copy Markdown
Member

Requested by Joep · project thread

Before: On Android Chrome, opening the keyboard in the meeting chat (and in the More menu's filter) left the panel extending under Chrome's bottom toolbar. The composer and Send button were cut off and the panel header scrolled out of view.

After: The viewport meta asks for interactive-widget=resizes-content, so Chrome shrinks the layout viewport to the visible area above the keyboard and toolbar, and 100dvh layouts fit.

Chrome by default only shrinks the visual viewport, so 100dvh and the fixed panel stayed full height. The Android WebView ignores the hint, so useKeyboardInset stays as the fallback there.

Not reproduced locally: headless Chromium cannot emulate Chrome Android's toolbar and keyboard, so this needs a check on a real device. Screenshots of the reported behaviour are in the thread.


Generated by Claude Code

Chrome only shrank the visual viewport, so the meeting chat composer and
the More menu stayed under the keyboard's toolbar and were cut off.

joepio commented Oct 3, 2026

Copy link
Copy Markdown
Member Author

The Downstream (atomic-saas builds against this commit) check is red for a reason outside this PR. Its "Pick the atomic-saas branch" step failed with gh: Bad credentials (HTTP 401) when it called the GitHub API for atomic-saas, so the saas build never started. The PR only changes the viewport meta and a comment. The main CI job is still running.

Next step is for whoever owns the token that job uses to renew it, then re-run the job. No fix to port.


Generated by Claude Code

joepio commented Oct 3, 2026

Copy link
Copy Markdown
Member Author

Main (Mancave) / CI also failed: the Playwright progress line shows a handful of red e2e tests early in the run. The job log is cut at 4 MB, so it never reaches the list of failing tests, and I can't say which ones. This PR only adds a viewport meta hint and a code comment, and desktop Chromium ignores that hint, so I don't expect it to be the cause. I have not confirmed that.

I've re-run the failed jobs once to see whether the failure repeats. If it does, I'll look at the failing tests from the run's artifacts. Downstream is red for the bad-credentials reason in my earlier comment and will stay red until that token is renewed.


Generated by Claude Code

joepio commented Oct 3, 2026

Copy link
Copy Markdown
Member Author

The re-run is red again, now with a cause I can name. data-browser test failed 11 of 1555 tests, all in TableCell.test.tsx (10) and TableViewTabs.test.tsx (1). The first of them hit Test timed out in 5000ms (7.6 s to render), and the rest then failed with No cell rendered. The run took 388 s with 3327 s of import time, so the runner was heavily loaded.

This PR doesn't touch those files, which render table cells and view tabs. It only changes index.html and a comment, so I don't think it causes this, though I haven't proven it. Because the unit stage failed, the e2e suite never ran in this attempt. I've used my one re-run, and I know of no fix for this to port.

It looks like a timing problem in those two unit tests under load. That is for the CI owners to look at, and re-running on a quieter runner would show whether it clears.


Generated by Claude Code

joepio added a commit that referenced this pull request Oct 5, 2026
… chat, search, CI) (#2060)

Combines #1530, #1638, #1663, #1835, #1869, #1896, #1898, #1911, #1912, #1913, #1914, #1918, #1919, #1929, #1931, #1949, #1958, #1959, #1961, #1963, #1968, #1975, #1981, #1985, #1988, #2038, #2041, #2047 and #2050, tested in one full Mancave run on the exact head (run 4999, green).

joepio commented Oct 5, 2026

Copy link
Copy Markdown
Member Author

Landed on develop in the combined batch #2060 (95707c0), tested together in one full run. Closing as included.


Generated by Claude Code

@joepio joepio closed this Oct 5, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant