3 Commits
Author SHA1 Message Date
thepraandClaude Opus 5.5 9d662f702b The feed memory test counts the deletions it was told of
Build / Build (push) Successful in 30s
Deploy / decepub.thepra.dev (push) Successful in 1m33s
In one full run (241 of 242 passing) the home feed came back from a thread with fewer posts than it was left with, and
passed three times alone. The feed takes a post out at once when the stream says it was deleted, and earlier tests do
delete posts in that persona's home, but PrivaPub's log shows no federated Delete at that minute, so the cause is not
proven. The test now allows for the deletions the stream reported during the trip and says the three numbers when it
fails; with none reported, a shorter feed still fails, and coming back still may not read the home feed again.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LsXgEaXee4GCU1hwYgPJXw
2026-10-07 02:55:30 +02:00
thepraandClaude Opus 5.5 b66d7cf545 Tests click exactly what they mean
A card's time is a "Post" button now (its nerd stats open with a tap), so the poll test submits through the composer's
own submit button instead of the last button named Post; the feed-memory test opens a thread from the Expand button,
not from the first expand/ link, which on a desktop is now the phone-only Thread entry. The Back flow presses Back with
history.back(): a Back held to close an overlay loads nothing, which GoBackAsync waited for.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LsXgEaXee4GCU1hwYgPJXw
2026-10-07 00:20:56 +02:00
thepraandClaude Opus 5.5 23dbc68fb4 Coming back to the feed shows it as it was left
Leaving the feed keeps it in Services/FeedMemory.cs (posts, waiting posts, cursors and scroll; ten minutes, for the same
feed of the same persona), so coming back from a thread or a profile shows it again at once, older pages included,
instead of reading it anew; newer posts come behind the "new" button and through the stream as before.
tests/e2e FeedMemoryTests goes to a thread and back without a new read of the home timeline; the whole suite passes
(123, 5 skipped).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LsXgEaXee4GCU1hwYgPJXw
2026-10-06 01:21:41 +02:00