[pull] develop from marktext:develop - #115
Merged
Merged
Conversation
…+ load) (#4861) `loadLanguage` invoked Prism's `getLoader().load()` without a Promise chainer, so the loader fired a dependent component's import without awaiting its dependency. A component whose grammar extends another (`cpp` extends `c`) could therefore evaluate before the dependency was registered: `Prism.languages.extend('c', …)` ran on `undefined` and threw "Cannot set properties of undefined (setting 'class-name')". That thrown error also left the load promise unresolved, hanging the caller — which surfaced as an intermittent 5s timeout in languageAlias.spec.ts on CI. Pass Prism's `series`/`parallel` chainer (its documented `Promise#then` / `Promise.all` hooks) so a dependency is imported and registered before its dependent. Await the composed loader promise and collect statuses directly instead of the ad-hoc deferred array (which only worked because the no-chainer loader happened to invoke callbacks synchronously). Reproduced the race locally (2/15 runs) and confirmed the fix eliminates it (0/30); added loadLanguageDependencyOrder.spec.ts pinning the dependency order (3/15 failing before the fix, deterministic after). Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…th (CommonMark) (#4856) * feat(muya): add firstWordOfInfo helper for code fence info strings Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * refactor(muya): make meta.lang hold the full code fence info string Store the whole info string verbatim on meta.lang and serialize it as-is; the language is derived as its first word. Removes the redundant meta.info field and its staleness guard (the single field is now the source of truth). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * refactor(muya): derive code highlight language from the info string's first word Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(muya): let the code block language input edit the full info string Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * refactor(muya): centralize code fence language derivation (review cleanup) Applies /simplify review feedback to the info-string refactor: - Derive the highlight/tokenize language once in `CodeBlockContent._lang` (returns the info string's first word) instead of at each call site. This fixes two consumers the scattered approach missed — the tabHandler markup test and the Firefox-compat backspaceHandler both read the raw field and silently mis-derived the language for a multi-word info string. - Move `firstWordOfInfo` from block/commonMark/codeBlock into utils, so the state layer no longer reaches into the block tree for it (removes a reverse state -> block runtime import). - Simplify `lang: isFenced ? info : lang` to `lang: info` in `_buildCodeState` — indented blocks have no info string, so both arms are equal. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
) (#4854) The "/" quick-insert menu and the block "turn into" front menu labelled their h1-h6 section "HEADERS" with items "Header 1".."Header 6". In CommonMark/HTML terms those are headings; "header" means a page header or front matter, so the wording was incorrect (only in English — the other shipped locales already translated these as heading/title). Rename the section (headers -> headings) and item titles (Header N -> Heading N) in the menu config, and rename the matching i18n keys across all locale files so every translation keeps resolving (their translated values are unchanged). The internal block label ('atx-heading N') and icon paths are untouched. Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…#4853) The image-path auto-complete helper calls fs.watch() on the directory of the path being typed, with no error handling. fs.watch throws synchronously for directories the OS can't watch — notably UNC / \\wsl.localhost network paths on Windows (EISDIR). The throw happened inside the fs.readdir callback of searchFilesAndDir, so it escaped as an uncaught exception and surfaced to the user as the generic "Unexpected error occurred in the main process" dialog. Wrap fs.watch in try/catch and attach an 'error' listener to the watcher so auto-complete silently degrades to "not watching" that directory instead of crashing the main process. Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
) * fix(desktop): save files atomically via temp-file + rename (#3786, #3828) writeFile() wrote directly to the destination with fs-extra outputFile, which truncates the existing file before streaming the new content. A crash, BSOD, or power loss mid-write therefore left the user's document truncated to 0 bytes (one report attached a 44393-byte file that was entirely NUL). This is the data-loss the maintainers' own TODO in writeMarkdownFile flagged as "safeSaveDocuments". Write to a temp file in the same directory and rename it over the target instead. The rename is atomic on a single volume (guaranteed by keeping the temp file in the same directory), so an interrupted write can only corrupt the throwaway temp file, never the existing document. outputFile still recreates any missing parent directory first, preserving #3509. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(desktop): make the atomic save durable against power loss (#3786, #3828) Addressing review: the temp-file + rename change fixed the application-crash case but not the power-loss / OS-reboot case the issues actually describe. A bare rename is namespace-atomic but not data-durable: after a power loss the new directory entry can point at an inode whose data blocks were never flushed — a full-length, zero-filled file, exactly the 44393-byte all-0x00 artifact in #3786. It also dropped the in-place write's mode/owner preservation, symlink follow-through, and unique temp naming. Delegate to write-file-atomic (fsync-before-rename), which closes the power-loss window and restores those behaviors: it stat/chmod/chowns the temp to match the target, writes through a symlink via realpath, and uses a collision-free temp name. ensureDir keeps the #3509 missing-directory behavior. Tests now cover overwrite/buffer/missing-dir and permission-mode preservation. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * chore(desktop): type write-file-atomic via @types instead of a local shim @types/write-file-atomic@4.0.3 declares the same public API write-file-atomic 7 exposes (its signature has been stable since v4), so drop the hand-rolled module shim for it. Every writeFile caller passes a BufferEncoding or nothing, so narrow the options parameter from WriteFileOptions to BufferEncoding — that also sidesteps the Node Mode (string | number) vs @types mode (number) mismatch cleanly. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
… loss (#4864) * build(desktop): add write-file-atomic dependency Adds write-file-atomic (+ @types) for durable, atomic file writes. Used by the crash-recovery buffer store; the document save path adopts it separately in #4852. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(desktop): make crash-recovery buffer writes durable against power loss writeBufferStoreFile persists unsaved tab content (the crash-recovery buffer) with a temp-file + rename. That is namespace-atomic — safe against an application crash — but omits fsync, so a power loss / OS reboot can still commit the new directory entry while the data blocks were never flushed, leaving a truncated or full-length zero-filled buffer. This is the same gap the document save path had (#3786). Write through write-file-atomic's sync API (fsync before the atomic rename), which also removes the hand-rolled temp/rename/cleanup and the writeSequence counter (its unique temp naming is handled internally). 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.
1 participant
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 : )