Repository navigation
chore: simplify escaping and release 0.5.0 - #42
Merged
Merged
Conversation
A character that XML 1.0 forbids, such as U+0001, in a Blog post's title, body or description used to fail the whole build when the Atom feed was written. The shared content rules now reject it for that post, so the post is skipped (exit code 2) and the draft lint reports it too. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A Theme that extends Quiet may replace header.html without the appearance button. site.js then threw before it built the table of contents and the copy-code buttons. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A live site builds with the 0.4.0 Action; a repository made from the template has not been tried end to end yet. escaping-template now matches starter/. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- AGENTS.md: Themes can come from GitHub and extends can name one; templates run in Jinja's sandbox; the file list matches src/. - Issue content contract: a bad Blog or Idea Issue is skipped, not a failed build; the oldest About wins; tag keys may be Unicode; paths are defaults that pages can move. - ADR 0007 and 0008 say which later ADRs amend them. - The draft docs agree with the code: the Issue body always has a front matter block. The lint skill names /assets/. - testing.md lists the remote Theme, Atom and draft test owners and drops the finished rebuild plan. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Source Serif 4 and --serif were left behind when the post body moved to the sans stack; browsers never downloaded the font. - Manrope ships as WOFF2 (31 KB) instead of TTF (95 KB); same glyphs. - Remove declarations that repeat inherited or global values or can never apply: text-decoration: none on links (the global a rule has it), a hover colour equal to a:hover, color: inherit on a span, a max-width wider than <main>, and a checkbox rule the sanitizer makes unreachable. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- ThemeConfig.local_path and RouteRegistry.route_for_url had only tests as callers. - parse_front_matter's raising mode for unknown fields: the compiler always collects them and reports each one. - SiteCompiler(output_staging=...) was never passed. - OptionSpec copied the theme.yaml option model field by field; the model is now OptionSpec. - RouteRegistry rejected .html in page paths, which every path validator already rules out; the output path check stays. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Old Config fields and a theme.yaml pages: key are now plain unknown fields, and an old Theme API gets the one api: 4 message. The Theme guide keeps the tables of where each setting went. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
.gitignore was the generic Python template plus old site paths; the short version ignores the same files in this checkout. The independent test Theme never used mark.svg. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The feed builder re-checked every post for timestamps without a time zone and for characters XML forbids, and the renderer checked again. Blog posts are already checked when they are read, so the feed is now one function that renders the SiteModel, with the same bytes as before. The remaining checks moved to where the text comes in: the site title, author and description are checked with the Config, and an Issue time without a zone is refused when the Issue is read. The renderer still refuses to write a feed XML cannot parse. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Blog archive and the tag pages were each a module plus a model module, and the tag result carried a diagnostics list that was always empty. They are now two functions in site_builder.py, and ArchivePage and Tag live with the rest of the SiteModel. ArchivePage loses two URL properties nothing read; Tag keeps them, since Themes use them. The ContentCompiler already sorts Blog posts and Ideas newest first, so the archive, tags, feed, search index and render context stop sorting them again. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…rest Blog and project slugs were checked by three copies of the same pattern; routes.SLUG is now the only one. The reserved Blog slug "page" was checked separately by the ContentCompiler and the local draft check; it is now part of the shared content rules, with the same error. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The module described publication four times over, in the module, class and method docstrings. It now says it once. Four one-line wrappers around the ownership check and the registries are gone; every path is still checked before it is renamed or deleted, with the same errors. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
structlog was used for five lines, printed to stdout next to the GitHub Actions annotations, and three of them repeated an error or warning the build already reports. Those three are gone. The downloaded Theme's commit and the traceback of an unexpected crash now go through logging, which the escpe command prints to stderr as plain lines. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Quiet's widths, font sizes, line heights, borders and palette were each asserted to the pixel or the RGB value, so any restyle failed the suite without anything being broken. Those assertions are gone. The tests still check what a reader relies on: pages stay centered and do not overflow or overlap, controls are large enough and show focus, the layout does not shift when scripts load, Mermaid text keeps its contrast, reduced motion stops movement, and code keeps its text and scrolls. The two browser launches now share one helper, and the fixture that runs a test in both Chromium and WebKit is named for that. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Two tests built Issues out of MagicMock and mostly checked the mocks: that no field was a MagicMock and that no mocked write method was called. The test with real PyGithub objects now compares the whole snapshot, and the local GitHub stand-in, which answers only GET, already fails any write. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Merge the API 3 and API 2 migration notes into one table set that maps straight to API 4, point every link at it, and give the README a single upgrade path. Drop the claim that old 0.1 fields tell you where they moved; since the hints were removed they are reported as unknown fields. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
AGENTS.md now points to the testing guide instead of repeating its test budget. The guide folds the "minimum set" list into the budget, the domain doc drops advice for files that already exist, and the build diagram shows that the SiteBuilder now pages the Blog and builds tags. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- CI: checkout 7.0.1, setup-python 7.0.0, setup-uv 10.2.0. - Action and CI: uv 0.12.17, the newest release setup-uv 10.2.0 has a checksum for, so the download is still verified. - Starter: deploy-pages 5.0.1. - Mermaid 11.17.2 from the npm package (integrity checked); 12.0 is a new major and waits. .gitattributes still named the bundle's pre-0.2 path; it now marks the current file binary, so whitespace checks skip it. - Pygments 2.21.0: the regenerated token rules are identical, so only the header of syntax.css changes. - uv lock --upgrade. ty 0.0.84 could not infer the sort key lambda in _newest_first, so it uses attrgetter; one ty: ignore is no longer needed. - Drop the ruff ignore for the deleted tests/test_atom_feed.py. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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.
问题与改动
仓库变臃肿了：同一件事有多处实现，有的代码没人用，测试钉死了 Quiet 的具体尺寸和颜色，文档重复，依赖和 Action 版本也旧了。这个 PR 做减法，Config 用法不变，站点输出除 Mermaid 升级和 sitemap 里几条 URL 的顺序外不变（例外见“风险与兼容性”）。
atom_feed.py478 → 约 70 行）。站点标题、作者、描述里有 XML 不允许的字符，现在在读 Issue 之前就报 Config 错误。page的检查和其他内容规则放在一起。docs/agents/testing.md一处。不在范围内：新功能、Theme API 变化、Mermaid 12。
验证结果
uv lock --check、CI=true uv run pytest -q -ra、ruff check、ruff format --check、ty check、整个分支的git diff --check）：693 passed，无跳过，全部通过。测试数比 main 少，是因为删了重复和基于 mock 的测试。sitemap.xml——81 个 URL 完全相同，只是 Blog 分页的 3 个 URL 排到了 Projects 后面（分页并进 SiteBuilder 后注册顺序变了；sitemap 的顺序没有含义）。TypeError: Load failed在 v0.4.0 的产物上一模一样，不是这次引入的。风险与兼容性
继承 Quiet 且链接了
manrope-bold.ttf或source-serif-4.ttf的 Theme 需要自带这两个文件。旧字段（0.2/0.3 之前的写法）现在报“未知字段”，不再提示新位置；在 0.4.0 上它们本来就会报错，能正常构建的站点不受影响。
进度日志改成 stderr 上的普通文本行。
回滚：站点把 pin 改回
@v0.4.0。已遵循维护指南与测试策略，说明适用检查的实际结果
未弱化内容授权、sanitizer、路径 containment、artifact validation 或失败保旧产物
公共行为变化已同步契约与迁移说明；远程发布或站点变更有单独授权（如适用）
🤖 Generated with Claude Code