vole

Clone

Read only, no credentials:

https://rodent.software/git/vole.git

Write over SSH with a registered key:

git clone ssh://git@0.0.0.0:2222/vole.git

A free and open-source forge that lives in your Git repo.

Serve repositories from the root of the URL space

ed03b3d1fa16ef6b55e2773aaf10e3be13282c3a · Keenan Nemetz · 4h ago · parent d37d301f

Serve repositories from the root of the URL space

A repository is now /<name> and its Git transport /<name>.git, so the URL a
person opens, the URL a clone uses, and the Go module path are the same name.
The forge's own pages moved under /-/ (admin, accounts, the directory, the
feed, sign-in, assets), a prefix no repository name can start with: names are
[a-z0-9...] and cannot begin with "-". That leaves the router with two disjoint
spaces and removes the /r/ and /git/ prefixes, which existed only to keep
repositories away from the forge's own paths.

Routes: repoRoutes lose /r/; global, auth and admin routes gain /-/; assets
move to /-/static/ because a subtree pattern like /static/ cannot coexist with
one-segment-wildcard repo routes — Go's mux reports them as conflicting
neither-more-specific patterns. The catch-all still renders the styled 404.

Transport: githttp.Route parses /<name>.git/<rest> instead of a /git/ prefix,
and forgepaths.go hands such requests to the proxy before the mux, because a
wildcard fills a whole segment and cannot match "<name>.git". LFS stays on the
mux (<name>.git/info/lfs/...), and the Go module discovery probe moves to the
same pre-router so a package path under a public repository is answered even
when its first segments name a page tab.

Names: repo.ValidateName now refuses the four root paths the forge must keep —
static, robots.txt, sitemap.xml, sitemap-pages.xml. Everything else is an
ordinary repository name, including admin, u, and activity.

Docs and the packaged URLs follow the new layout; robots.txt, the sitemaps and
the instance activity feed point at the new paths. No aliases: the old /r/ and
/git/ URLs are gone.

Verified with ./vole workflow check and ./vole workflow run verify.

server/read_limits_test.go +4−4

@@ -34,7 +34,7 @@ func TestLargeBlobIsBounded(t *testing.T) {
	34	34
	35	35	// The inline view is a preview: the first MiB, and the file's real size stated
	36	36	// rather than the size of the preview.
	37		code, page := f.get("/r/" + f.repo + "/blob/main/big.txt")
		37	code, page := f.get("/" + f.repo + "/blob/main/big.txt")
	38	38	if code != http.StatusOK {
	39	39	t.Fatalf("blob page = %d", code)
	40	40	}
@@ -52,7 +52,7 @@ func TestLargeBlobIsBounded(t *testing.T) {
	52	52
	53	53	// The raw download carries all of it, streamed, with a length the client can
	54	54	// trust before the first byte arrives.
	55		resp, err := http.Get(f.http.URL + "/r/" + f.repo + "/raw/main/big.txt")
		55	resp, err := http.Get(f.http.URL + "/" + f.repo + "/raw/main/big.txt")
	56	56	if err != nil {
	57	57	t.Fatal(err)
	58	58	}
@@ -73,7 +73,7 @@ func TestLargeBlobIsBounded(t *testing.T) {
	73	73
	74	74	// A README past the cap is named and measured instead of rendered: silently
	75	75	// dropping it would look like a bug in the forge.
	76		code, overview := f.get("/r/" + f.repo)
		76	code, overview := f.get("/" + f.repo)
	77	77	if code != http.StatusOK {
	78	78	t.Fatalf("overview = %d", code)
	79	79	}
@@ -85,7 +85,7 @@ func TestLargeBlobIsBounded(t *testing.T) {
	85	85	}
	86	86
	87	87	// Small files are unaffected.
	88		if code, body := f.get("/r/" + f.repo + "/blob/main/small.txt"); code != http.StatusOK \|\| !strings.Contains(body, "small") {
		88	if code, body := f.get("/" + f.repo + "/blob/main/small.txt"); code != http.StatusOK \|\| !strings.Contains(body, "small") {
	89	89	t.Fatalf("small blob = %d", code)
	90	90	}
	91	91	}