| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
| |
Detect path tokens in prose (must contain a "/") and inline <code> spans
(matched verbatim, so bare Makefile/README link too) and turn them into
links to the tree/ (or about/ for *.md) view — but ONLY when the token
resolves to a path that actually exists in the repo, which we learn from
`git ls-tree -r -t` at the default branch. <pre> code blocks, existing
<a> links (no nested anchors) and tag attributes are never touched;
tokens resolve relative to the rendered file dir or the repo root.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
The list-interrupt preprocessor inserted a blank line before any bullet
that followed a non-item, non-4-space-indented line. When a list item
wrapped onto a 2-space-indented continuation line, the *next* item looked
like a paragraph->list transition, so a blank got injected between items
— turning a tight list into a loose one (every <li> wrapped in <p>), as
seen in docs/internals.md "Writing programs in C".
Track list state and only interrupt an actual paragraph: a bullet that
continues an already-open list (or its indented continuation lines) no
longer triggers a blank-line insert. The original paragraph->list
interrupt and the "1985." ordered-list guard are unchanged.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Links to markdown files now point at /<repo>/about/<path> (rendered
through the about-filter) instead of /<repo>/tree/<path> (raw blob), so
clicking a README link to another .md shows rendered markdown.
Also resolve relative links against the rendered file's own directory:
about-formatting.sh now forwards the filename cgit passes ($1) to
md2html as argv[1], and md2html uses dirname(argv[1]) as the base. This
makes links *inside* a rendered subpage (docs/foo.md -> ./bar.md,
../README.md, images) resolve correctly. Non-markdown files and
directories still go to tree/; images to plain/; about/ links carry no
?h= since the about view renders against the configured readme ref.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The image-rewrite patch only fixed relative <img src>; relative <a href>
links (e.g. [docs/x.md](docs/x.md)) were still resolved by the browser
against the about/summary page URL, not the repo root, so they 404d.
Rewrite relative <a href> to absolute cgit URLs the same way:
in-repo path -> /<repo>/tree/<path>?h=<branch>
../sibling -> resolved against the repo URL location (/<sibling>)
Fragments are preserved; anchors, absolute, scheme and query-only URLs
are left untouched.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Python-Markdown, unlike CommonMark/GitHub, refuses to start a list that
immediately follows a paragraph line without an intervening blank line, so a
very common README shape renders as one run-on paragraph:
Passing:
- foo
- bar
Add a preprocessor that inserts the blank line when a bullet (- * +) or a
1./1) ordered marker follows a paragraph line, restoring GitHub behavior.
Only '1' interrupts for ordered lists (matching CommonMark) so prose like
'... the year\n1985. was ...' is not turned into a list. Fenced/indented
code and lines already inside a list are left untouched.
|
| |
|
|
|
|
|
|
|
| |
A README's relative image paths (screenshots/foo.png) are repo blobs that
cgit only serves via its plain command, but a browser resolves a relative
<img src> against the about-page URL, so the image 404s. Rewrite relative
<img src> to an absolute /<repo>/plain/<path>?h=<branch> URL using the
CGIT_REPO_URL / CGIT_REPO_DEFBRANCH environment cgit gives the about-filter.
Absolute so it also works with ui-summary readme rendering.
|
| |
|
|
|
|
|
| |
This addressed a non-existent background image and made the element
invisible. Drop the style and use something sane.
Signed-off-by: Christian Hesse <mail@eworm.de>
|
| |
|
|
|
|
|
| |
This allows for cleaner nesting semantics and matches github more
closely.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
|
| |
|
|
|
|
|
| |
This makes the markdown filter generate anchor links for headings.
Signed-off-by: Christian Hesse <mail@eworm.de>
Tested-by: jean-christophe manciot <actionmystique@gmail.com>
|
| |
|
|
|
|
|
|
|
| |
Otherwise we get the classic Python UTF-8 errors, and the text is all
out of order. While we're at it, switch to python3 so we only have to
support one set of oddball semantics.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Suggested-by: Daniel Campbell <dlcampbell@gmx.com>
|
| | |
|
| |
|
|
| |
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
|
| |
|
|
| |
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
|
| |
|
|
| |
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
|
|
|
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
|
