[pull] develop from marktext:develop#96
Merged
Merged
Conversation
) Selecting a word and opening the find bar is meant to prefill the find input with the selection. The prefill is driven reactively by `watch(searchMatches)`, which mirrors the editor's current selection into the input. But opening the bar steals focus from the editor, and the engine emits a spurious selection-change for the reset selection. Because the watch is async, that change coalesces with the intended one and can cancel it, leaving a stale value in the input (the regression guard observes a single leading character, e.g. "T"). Fix in two parts: - Seed the input synchronously from the current selection when the bar opens (`prefillFromSelection`), so the prefill can't lose the race with the focus-steal selection-change. - Once open, the bar owns the query: ignore editor selection-changes in the `searchMatches` watch while `showSearch` is true, so the focus-steal change can't clobber the prefilled/typed value. Covered by test/e2e/search-prefill-from-selection.spec.ts (the race reproduces under the Linux/xvfb CI environment).
* fix(muya): preserve nested-list indentation in RTL mode (#4673) List containers used physical `padding-left` and the task-list checkbox used physical `left`, neither of which follows `dir="rtl"`. In RTL the padding stayed on the left while markers flipped to the right, so every nesting level collapsed onto the same right-hand axis. Switch to the logical `padding-inline-start` / `inset-inline-start` equivalents. These are identical to the physical properties under LTR (no visual change) and mirror to the right under RTL, restoring the visual hierarchy of nested ordered, bullet, and task lists. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(desktop): stop down-compiling CSS logical properties to LTR physical `postcssPresetEnv({ stage: 0 })` enabled the `logical-properties-and-values` transform (postcss-logical), which rewrites `padding-inline-start` / `inset-inline-start` to hard-coded LTR `padding-left` / `left` at build time. That silently defeated the list-indentation logical properties under `dir="rtl"` — the served CSS always resolved to the left regardless of text direction. Electron ships Chromium, which supports logical properties natively, so the down-compile is unnecessary. Disable just that feature; everything else in preset-env is unchanged. Verified in the Electron 42 Chromium: nested lists now resolve `padding-right: 30px` per level under RTL, stepping the hierarchy in from the right edge (#4673). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…4670) (#4674) The math renderers imported the chemistry extension via `katex/dist/contrib/mhchem.min.js`. That deep path falls through the katex exports map to the UMD bundle, whose internal `require("katex")` resolves to the CommonJS build (`dist/katex.js`) — a different module instance from the ESM `import katex from 'katex'` (`dist/katex.mjs`) the renderers actually call. mhchem patched the CJS instance while rendering used the ESM one, so `\ce{...}` stayed an undefined control sequence and fell back to "Invalid Mathematical Formula". Import the ESM build `katex/dist/contrib/mhchem.mjs` instead, which extends the same deduped instance. Also add the import to the HTML-export render path in marked/extensions/math.ts, which was missing it entirely. Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…4671) (#4672) A well-formed `[text](url)` (or reference) link was dropped whenever its destination URL also matched the GFM extended (bare-URL) autolink rule and was immediately followed by more content — e.g. a CJK comma `[t](https://x)、`, trailing text `[t](https://x)foo`, or simply the first of two links on one line. The rendered output showed the raw markdown with the link styling bleeding into the following text. Root cause: `tryLink`/`tryReferenceLink` validated the tentative link with `lowerPriority(..., validateRules)`, the full inline-rule set. Two of those rules can match a span that runs past the link and falsely veto it: - `auto_link_extension` matches the bare destination URL plus any trailing punctuation/text, so it extends beyond the link's closing paren. - `link`/`image` themselves use a greedy `.*` destination and (before `correctUrl` runs) match across to a later link on the same line. Per CommonMark §6.6 only code spans, raw HTML tags and `<...>` autolinks bind more tightly than a link, so only those may defer it (the documented CM 0.29 examples 520/521). Introduce `linkValidateRules` containing exactly those three and use it for both link validators. Extended autolinks and angle autolinks in plain text, and the code-span/HTML-tag precedence cases, all still behave. Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…hange (#4681) * feat(menu): add AppMenu.updateKeybindings to rebuild menus from current bindings Rebuild every window menu (editor + macOS settings) so accelerators are re-read from the live keybinding map, re-applying the active window's menu. Editor menus preserve their runtime toggle state. This is the capability the keybinding-save flow needs to refresh the menu bar without a restart (#3998). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(keybinding): sync menu bar and command palette on keybinding save After persisting user keybindings, rebuild the application menu via AppMenu.updateKeybindings() and push the refreshed accelerator map to editor windows so the command palette updates too. Previously the new bindings only took effect after an app restart, so the menu bar and command palette kept showing the old shortcuts (#3998). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
) * fix(muya): guard detached block in code language selector (#4654) Selecting a language from the code-block language picker crashed the renderer when the picker's target block had been detached from the document while the picker stayed open (e.g. the code block was converted back to a paragraph via Backspace at offset 0 of the code body, orphaning its language-input child). `selectItem`'s `block.text = name` runs the Content text setter, which re-computes an OT path for the edit operation. For a language-input that path walks up to its code block, whose `path` getter dereferences its own (now null) parent — throwing `Cannot destructure property 'path' of 'this.parent'`. The retired muyajs engine hit the same class at `ContentState.updateCodeLanguage` (`reading 'functionType'` of null). Bail out of `selectItem` when `block.outMostBlock` is null — the block (or any ancestor) is no longer connected to the document root — before mutating its text/lang. `outMostBlock` walks the full parent chain, so it catches the detached-grandparent case a shallow `block.parent` check misses. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(muya-e2e): cover #4654 language selection on a detached code block Drives the real CodeBlockLanguageSelector float in Chromium: opens the picker on a code block nested in a list, detaches the block, then clicks a language item and asserts no uncaught renderer pageerror. Fails on the pre-fix engine with `Cannot destructure property 'path' of 'this.parent'`. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(muya): auto-hide code language picker when caret leaves its block The language picker only hid on a non-matching query or a document click, so moving the caret out of the language input (e.g. Left/Right arrow) left it open over a stale target. Subscribe to `selection-change` and hide the picker when the caret's anchorBlock is no longer its target block — the same self-perceiving pattern InlineFormatToolbar uses. This also closes the window that let the target block be detached from under the picker (#4654); the selectItem outMostBlock guard remains as defense in depth. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <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 subscribe to this conversation on GitHub.
Already have an account?
Sign in.
2 participants
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.
See Commits and Changes for more details.
Created by
pull[bot] (v2.0.0-alpha.4)
Can you help keep this open source service alive? 💖 Please sponsor : )