Text chat 31: Threads UI (#281) #393
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "textchat-31-ui-threads"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
What this adds
screenshare/settings_menu.rs), both the full menu and the compact live one. These rows come right after the sound checkbox, for a screen share or a share that has not started yet, while sound is on andStreams::audio_targetsis notUnavailable:audio_sources::menu_items, the label fromaudio_summary, and the picks fromtoggle_include/toggle_exclude/pick_system/pick_none.LinuxAudioCapturecheckboxes in the original's order and wording. Each one sendsVoiceSettingChange::LinuxAudioCapture(..), and the store turns off the other switch of the Ignore-devices / Device-selection pair.UserAction::RefreshAudioSources, and so does "Refresh audio sources".screenshare::MessagegainsPickSystemAudio,PickNoAudio,ToggleIncludeSource,ToggleExcludeSource,RefreshAudioSourcesandLinuxAudioCapture. The labels ofEntry::ToggleandEntry::Submenuchange toString, because the source names and the summary are built at run time.voice_video/devices.rs): newDefaultPlatform::{Linux, Windows, MacOs}withcurrent(), anddefault_label(platform, endpoint), which gives "Linux default (X)", or plain "Linux default" when there is no default device. The default device is listed first, then the others by name, ignoring case.DEFAULT_LABELis removed.docs/parity/screenshare.mdis now done, and an "Audio settings" row is added.Tests
crates/fluxer-ui/tests/stream_audio_sources.rs:audio_sources_submenu_only_for_screen_share_with_sound,audio_sources_hidden_without_pipewire,summary_in_the_submenu_label,include_and_exclude_groups_follow_the_mode,refresh_keeps_the_menu_open_and_asks_the_store,audio_settings_checkboxes_in_order_with_defaults,ignore_devices_and_device_select_toggle_each_other_in_the_menu.settings_voice.rs:default_option_names_the_platform_and_endpoint,devices_sorted_default_first_then_by_name, and the existing dropdown test updated.stream_menu_audio_sources,stream_menu_audio_settings,voice_devices_linux_default. The goldens were made on lavapipe in the Ubuntu 22.04 image.control_bar_sharing,stream_settings_menuandsettings_voicewere regenerated, because they now show the submenus or the new default label. All 35 screenshot tests pass on lavapipe, and also on this machine's GPU.screenshare_controls.rs(menu labels) andcall_view.rs(the caret now also sendsRefreshAudioSources).FLUXER_SCREENSHOT_TESTS=1 cargo xtask ciis green locally.Deviations and notes
NotAskedorLoading, as the original does while it loads. On Windows and macOS the backend answersNoneand they disappear, so they may show for one frame on the first open.StreamSettingsMenuContent.tsx). With the defaults, all seven show.textchat; the branch has origin/textchat (67da1c7) merged in (conflicts indocs/parity/screenshare.md, kept both the stereo row and the audio-sources rows, andcrates/fluxer-ui/README.md, kept the composer text plus the device-dropdown sentence).FLUXER_SCREENSHOT_TESTS=1 cargo xtask cigreen again after the merge.fixtures/golden/.Closes #343
🤖 Generated with Claude Code
Independent review: Text chat 31 (Threads UI)
Reviewed
git diff 656cd729..origin/textchat-31-ui-threads(22dcae2) cold.Checks
chip_click_opens_panel,panel_width_clamped(360/900/450, clamps 359→360, 901→900),create_pane_emits_create,sidebar_thread_rows_indented(1.75 rem indent, spine at 1.5625 rem) all exist incrates/fluxer-ui/tests/threads.rsand assert what the plan asks. Thethread_chipandthread_panelscreenshot tests and goldens are present. I looked at the goldens: chip ≤30 rem on--background-secondary-alt, panel at 450 px with the member list hidden.panel::EDITOR_ID), the create pane (name field; the server's default auto-archive), the sidebar rows under the parent, andPanelHeader::{Thread, ForumPost}(needed by Task 45) are all present. The Open-in-full-view, members and "..." buttons are stubs, as the Phase 5 note allows (Tasks 38 and 44).max(0.0)on spine geometry).cargo xtask ci: green (all steps passed). The screenshot step was skipped because this machine has no lavapipe ICD; the goldens are verified only by the Linux runner.Findings
screens/main/media.rs::wantedcollects inline emoji only fromlocal.messages.parsed_nodes(), so custom emoji in the side panel's messages are never requested. Author and reply avatars and reaction emoji are covered through the newshown_channels. This needs a follow-up issue (the implementer mentions it, but none is linked).messages.rs:279). The Phase 5 notes and Task 44's Interfaces reassign it toOpenThreadBrowser(Task 38), so deferring it is acceptable. Please note it in the PR body.SelectChannel(parent)+OpenThread). Spec §18.2.5 says the sidebar opens the full view, which is Task 44's. This is documented inthreads/mod.rs; Task 44 must change it.CreateThread. Pressing Enter twice before the THREAD_CREATE echo sends two create requests, which would create two threads. Suggest apendingflag, as the original's submitting state.thread_panel.pngthe narrower channel view clips the reply preview's@charliemention pill into a stray box (Task 24's reply_preview at narrow widths).Verdict: APPROVE. No blocking findings; local ci green. The minor items should become follow-up issues.
Independent re-review: Text chat 31 (Threads UI), after the textchat merge
I reviewed
git diff 9c96985..origin/textchat-31-ui-threads(dab3e62) without relying on the earlier review.Checks
crates/fluxer-ui/tests/threads.rshaschip_click_opens_panel,panel_width_clamped(min 360, max 900, default 450; 359→360, 901→900, 1000→900, 200→360),create_pane_emits_createandsidebar_thread_rows_indented(indent 1.75 rem = 28 px, spine at 1.5625 rem = 25 px). Thethread_chip_matches_goldenandthread_panel_matches_goldenscreenshot tests and their goldens are present. I checked the constants against the fork:ThreadPanelWidth.ts:7-9,MessageThreadChip.module.css(30 rem,--background-secondary-alt),SidebarThreadList.module.css, andDEFAULT_THREAD_AUTO_ARCHIVE_DURATION = 4320. All match.messages::State, keyed by the thread's id, and its own composer,panel::EDITOR_ID), the create pane, the sidebar rows, andpanel::PanelHeader::{Thread, ForumPost}, which Task 45 needs. The full-view, members and "..." buttons are stubs, which the Phase 5 notes allow (Tasks 38 and 44). The merge adapts cleanly to Text chat 25:messages::view_channel(.., &ComposerOptions, editor_id)andcomposer::view_with_id.unwrap_or_default, and the code only does string operations and clamped geometry.FLUXER_SCREENSHOT_TESTS=1 cargo xtask ciatdab3e62is green: fmt, clippy, deny, core-isolation, parity, tests and vendored tests all pass, and all 53 screenshot tests pass,thread_chipandthread_panelincluded.Findings
Closes #343. Merging as it is would close #343 (Audio 53, still open) and leave #281 open. Please replace the body with Text chat 31's, endingCloses #281, before merging.main::viewdrawscomposer::with_plus_menuonly fromlocal.messages.composer, the channel's, so toggling the panel's plus menu changes state that is never drawn. The implementer flagged this too. It needs a follow-up issue.media::wantedloads the attachments, stickers and embed images only forapp.ui.selected_channel, and the inline emoji only fromlocal.messages.parsed_nodes(). So a thread's pictures, stickers, embed media and custom emoji in the panel are never requested and stay as placeholders. Avatars and reaction emoji are covered byshown_channels. This is wider than the "emoji" finding of the first review. It needs a follow-up issue.CreateThreadtwice when it is submitted twice before the echo arrives;Per CLAUDE.md, these deferrals need
follow-upissues linked from the PR before merge.StartCreatehas no caller yet, so the create pane is unreachable in the app until Text chat 29's hover bar or menu sends it. That is expected, but Task 29 must wire it up.message_row::estimate_heightadds the chip's height whenevermessage.thread.is_some(), even whenchip::modelreturnsNone(a thread the user cannot see). The estimate is a little high; the effect is cosmetic.None of these is blocking: no interface is broken, no invariant is violated, no data is lost and CI is green.
Verdict: approved, provided the PR body is corrected (
Closes #281) and the follow-up issues above are filed before merge.