Repository navigation
Replies: 1 comment
|
Checked these against main. The folder counters as the total are deliberate, since counting rows on a large folder is slow, but three of the effects are real and I'll fix them on main:
I'd rather not adopt a broader window contract on top of that, so no need for a series here. |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Describe the bug
Follow-up beyond the category-total fix in #611: discuss a consistent row-window/pagination contract when cached folder badges differ from listable rows, a threaded page does not fill the viewport, or the last page disappears. These share the same boundary: how the list decides whether and where more visible rows exist.
Code inspection at main
8dbdc7982236404dbbb9de63dba769cfdb9ef701confirms that non-category totals still use folder counters and infinite loading is triggered from scrolling. The scenarios below are covered by synthetic fork regressions; I have not run the complete combined scenario on a clean-main instance, so the UI outcomes should be verified there before treating all of them as reproduced defects.To Reproduce
Use a disposable synthetic mailbox and a controllable API:
Expected behavior
Separate folder badge counts from listable-row boundaries; provide/consume bounded
hasMoreandnextOffsetsemantics (or an equivalent contract). Fill a short visible window while progress is possible, preserve already loaded windows on refresh, clamp invalid pages, and reject stale pages after navigation. In-flight refresh replay is already discussed in #612; category-tab total/unread UX is separately #597.Screenshots
None; synthetic fixtures only.
Desktop / Smartphone
Shared list logic. Synthetic fork UI checks passed Chromium and WebKit at desktop/mobile widths; installed iOS PWA behavior is not claimed.
Additional context
Public fork commit, PostgreSQL count regression, window tests, render regressions, and transport scenario. Fork CI passes. This does not reopen #611 or request merging the whole fork. Would this window contract fit upstream as a focused series?
All reactions