Repository navigation
fix: resolve 1.3.1 priority bugs (#350 #356 #357 #358 #359 #361 #368 #376) - #416
Merged
Merged
Conversation
The checksums job hashed only .dmg, .msi, .AppImage and .deb, so the Windows setup .exe and the Linux .rpm shipped without a published checksum. Also correct the release-body installer list, which named a non-existent x64-setup.msi and omitted the .exe and .rpm. Fixes #358 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The installation guide named Apple Silicon/Intel DMGs, a _x64-setup.msi and a lowercase kindling_*.deb that no release has shipped. Use the actual assets (universal DMG, _x64-setup.exe, _x64_en-US.msi, AppImage, Kindling_*_amd64.deb, Kindling-*-1.x86_64.rpm), add RPM install steps, drop the pick-your-architecture troubleshooting item, and fix the certutil and dpkg examples. Apply the same names to the release.yml releaseBody template. Fixes #357 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The README already headlines v1.3, has the v1.3 roadmap row, lists novelWriter import and export, editorial review, revisions, export profiles and writing goals, and places the revisions section above Contributing; the v1.3 README update did that. The one v1.3.0 changelog feature still missing was the automatic library backup on the first launch of a new version, so add it to New in v1.3 and the App table. Fixes #359 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Tauri's devtools feature was enabled unconditionally, so every webview in a release build allowed opening the Web Inspector, giving a JS console with invoke access to every IPC command. Debug builds still get devtools via debug_assertions. No code calls the devtools window APIs, so nothing needed cfg-gating. Fixes #361 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…omic create_chapter shifted later chapters and inserted the new one as separate autocommitted statements; reorder_beats validated and wrote positions in the same loop on the bare connection; promote_discovery_note_to_beat inserted the beat and deleted the note separately. A failure part-way left the outline half-updated (position gaps, half-applied orders, duplicated notes). Each now runs its writes in one unchecked_transaction. reorder_beats validates the whole order (including duplicate IDs) before writing anything. The snapshot metadata-failure cleanup from the issue was already in place. Fixes #376 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
export_workspace_document, get_export_prototype_document and save_export_prototype_html were synchronous Tauri commands, so they ran on the main thread and froze the window during large exports. Make them async like every other export command; names and arguments are unchanged. Fixes #350 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The title page shares the body's section and only had its header hidden by a distinct first-page header, so the first page of text was numbered 2. Start the section's page numbering at 0 when a title page is included, in both the classic and workspace DOCX exports, so the hidden title page is page 0 and the first page of text is page 1 as Standard Manuscript Format requires. Header text, fonts and margins are unchanged. Fixes #356 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…logout Dock > Quit, logout/shutdown and AppleScript `quit` call -[NSApplication terminate:]. In the pinned stack (tauri 2.10.3, tauri-runtime-wry 2.10.1, tao 0.34.5) tao's app delegate implements only applicationWillTerminate:, not applicationShouldTerminate:, so AppKit terminates immediately: no CloseRequested and no RunEvent::ExitRequested is raised (ExitRequested only fires for app.exit() and when the last window is destroyed). Handling ExitRequested alone would therefore not fix this; the debounced prose (500 ms) and synopsis (1 s) saves and the discard-on-failed-save prompt were skipped. Add applicationShouldTerminate: to the tao delegate class at startup (objc runtime, no new crates). It answers NSTerminateLater and emits `app-quit-requested`. The frontend runs the same quit() as the menu Quit: on success it calls exit(0), which arrives as ExitRequested with a code; the backend prevents that exit and approves the pending terminate so AppKit finishes it. If the save fails (discard prompt) or an update is installing, the frontend invokes the new cancel_app_quit command and the system quit is cancelled. AppKit replies are queued with performSelectorOnMainThread: (common modes) so they never run inside tao's event callback. Our own Quit path is unchanged: with no system quit pending, ExitRequested is not prevented, and a later exit after a reply goes straight through. Limits: - If the frontend does not answer within 10 s (hung or unloaded webview), the quit proceeds as before, so a quit can't hang forever. - A failed save during logout/shutdown cancels the logout; macOS shows that kindling interrupted it. That is intended (the discard prompt is shown) but the user must log out again. - macOS's own logout timeout for NSTerminateLater apps still applies. - If a future tao adds applicationShouldTerminate:, class_addMethod fails, a warning is logged and the old behaviour returns. - Runtime behaviour on a real Dock Quit/logout is not yet verified; only unit tests ran. Fixes #368 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
# Conflicts: # .github/workflows/release.yml
QUIT_REQUESTED_EVENT is only read by the macOS platform module, so clippy -D warnings on the Linux CI runner flagged it as dead code. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Priority fixes for the 1.3.1 patch release.
Changes
src-tauri/src/app_quit.rsaddsapplicationShouldTerminate:to the app delegate (tao 0.34 doesn't implement it), answersNSTerminateLater, and emitsapp-quit-requestedso the frontend runs the same flush/discard prompt as File > Quit. Ourexit(0)approves the pending terminate; a failed save calls the newcancel_app_quitcommand. 10 s fallback quits anyway if the webview doesn't answer.create_chapter,reorder_beatsandpromote_discovery_note_to_beatnow run in one transaction;reorder_beatsvalidates the whole order before writing. (Snapshot orphan-file cleanup was already onmain.)export_workspace_document,get_export_prototype_documentandsave_export_prototype_htmlare nowasync, so they no longer run on the main thread.pgNumType start=0on the section when a title page is present, so the first page of text is 1 (SMF).devtoolsfeature; debug builds still get it viadebug_assertions.*.exeand*.rpm.Fixes #350, Fixes #356, Fixes #357, Fixes #358, Fixes #359, Fixes #361, Fixes #368, Fixes #376
Testing
cargo test --all-features: 644/644.pgNumType/titlePgassertions, quit-gate state machine, App.svelte system-quit handler.Not verified at runtime
🤖 Generated with Claude Code