| 1 | // SPDX-License-Identifier: AGPL-3.0-or-later |
| 2 | |
| 3 | package repo |
| 4 | |
| 5 | import ( |
| 6 | "bytes" |
| 7 | "errors" |
| 8 | "fmt" |
| 9 | "html/template" |
| 10 | "net/http" |
| 11 | "path" |
| 12 | "strings" |
| 13 | |
| 14 | "github.com/go-chi/chi/v5" |
| 15 | "github.com/jackc/pgx/v5/pgtype" |
| 16 | |
| 17 | "github.com/tenseleyFlow/shithub/internal/auth/policy" |
| 18 | mdrender "github.com/tenseleyFlow/shithub/internal/markdown" |
| 19 | "github.com/tenseleyFlow/shithub/internal/repos/finder" |
| 20 | repogit "github.com/tenseleyFlow/shithub/internal/repos/git" |
| 21 | "github.com/tenseleyFlow/shithub/internal/repos/highlight" |
| 22 | reposdb "github.com/tenseleyFlow/shithub/internal/repos/sqlc" |
| 23 | "github.com/tenseleyFlow/shithub/internal/web/middleware" |
| 24 | ) |
| 25 | |
| 26 | // MountCode registers the code-tab routes: |
| 27 | // |
| 28 | // GET /{owner}/{repo}/tree/* |
| 29 | // GET /{owner}/{repo}/blob/* |
| 30 | // GET /{owner}/{repo}/raw/* |
| 31 | // GET /{owner}/{repo}/find/* |
| 32 | // |
| 33 | // The leading {ref} segment is variable-length (refs may contain `/`). |
| 34 | // chi's `*` wildcard captures the rest; we resolve ref + path inside |
| 35 | // the handler against the repo's known ref list. |
| 36 | func (h *Handlers) MountCode(r chi.Router) { |
| 37 | r.Get("/{owner}/{repo}/tree/*", h.codeTree) |
| 38 | r.Get("/{owner}/{repo}/blob/*", h.codeBlob) |
| 39 | r.Get("/{owner}/{repo}/raw/*", h.codeRaw) |
| 40 | r.Get("/{owner}/{repo}/find/*", h.codeFinder) |
| 41 | } |
| 42 | |
| 43 | // codeContext bundles the per-request data the code-tab handlers |
| 44 | // derive once at the top. Owner+repo come from chi; ref+path come from |
| 45 | // the wildcard, resolved against the repo's ref list. |
| 46 | type codeContext struct { |
| 47 | owner string |
| 48 | row reposdb.Repo |
| 49 | gitDir string |
| 50 | refs repogit.RefListing |
| 51 | allRefs []string |
| 52 | ref string // matched ref name (or 40-hex sha) |
| 53 | subpath string // path inside the ref, no leading slash |
| 54 | } |
| 55 | |
| 56 | // loadCodeContext does the resolve dance for tree/blob/raw/find. On |
| 57 | // any failure it writes the response and returns ok=false. |
| 58 | func (h *Handlers) loadCodeContext(w http.ResponseWriter, r *http.Request) (*codeContext, bool) { |
| 59 | row, owner, ok := h.loadRepoAndAuthorize(w, r, policy.ActionRepoRead) |
| 60 | if !ok { |
| 61 | return nil, false |
| 62 | } |
| 63 | gitDir, err := h.d.RepoFS.RepoPath(owner.Username, row.Name) |
| 64 | if err != nil { |
| 65 | h.d.Render.HTTPError(w, r, http.StatusNotFound, "") |
| 66 | return nil, false |
| 67 | } |
| 68 | refs, err := repogit.ListRefs(r.Context(), gitDir) |
| 69 | if err != nil { |
| 70 | h.d.Logger.WarnContext(r.Context(), "code: ListRefs", "error", err) |
| 71 | } |
| 72 | allNames := refNames(refs) |
| 73 | |
| 74 | rest := chi.URLParam(r, "*") |
| 75 | rest = strings.Trim(rest, "/") |
| 76 | segs := []string{} |
| 77 | if rest != "" { |
| 78 | segs = strings.Split(rest, "/") |
| 79 | } |
| 80 | if len(segs) == 0 { |
| 81 | // no ref → default branch root tree |
| 82 | ref := row.DefaultBranch |
| 83 | return &codeContext{ |
| 84 | owner: owner.Username, row: row, gitDir: gitDir, |
| 85 | refs: refs, allRefs: allNames, ref: ref, subpath: "", |
| 86 | }, true |
| 87 | } |
| 88 | ref, sub, ok2 := repogit.ResolveRef(allNames, segs) |
| 89 | if !ok2 { |
| 90 | // Fallback: if the first segment looks like a hex sha, accept it. |
| 91 | if len(segs[0]) == 40 && isHex(segs[0]) { |
| 92 | ref = segs[0] |
| 93 | sub = strings.Join(segs[1:], "/") |
| 94 | } else { |
| 95 | h.d.Render.HTTPError(w, r, http.StatusNotFound, "") |
| 96 | return nil, false |
| 97 | } |
| 98 | } |
| 99 | if !validateSubpath(sub) { |
| 100 | h.d.Render.HTTPError(w, r, http.StatusBadRequest, "") |
| 101 | return nil, false |
| 102 | } |
| 103 | return &codeContext{ |
| 104 | owner: owner.Username, row: row, gitDir: gitDir, |
| 105 | refs: refs, allRefs: allNames, ref: ref, subpath: sub, |
| 106 | }, true |
| 107 | } |
| 108 | |
| 109 | // codeTree renders the directory listing at <ref>:<subpath>. If the |
| 110 | // path turns out to be a blob, redirects to /blob/. README rendering |
| 111 | // for tree-roots is appended below the listing. |
| 112 | func (h *Handlers) codeTree(w http.ResponseWriter, r *http.Request) { |
| 113 | cc, ok := h.loadCodeContext(w, r) |
| 114 | if !ok { |
| 115 | return |
| 116 | } |
| 117 | h.renderRepoTree(w, r, cc) |
| 118 | } |
| 119 | |
| 120 | func (h *Handlers) renderRepoTree(w http.ResponseWriter, r *http.Request, cc *codeContext) { |
| 121 | kind, _, _, err := repogit.StatPath(r.Context(), cc.gitDir, cc.ref, cc.subpath) |
| 122 | if err != nil { |
| 123 | if errors.Is(err, repogit.ErrPathNotFound) { |
| 124 | h.d.Render.HTTPError(w, r, http.StatusNotFound, "") |
| 125 | return |
| 126 | } |
| 127 | h.d.Logger.WarnContext(r.Context(), "code: StatPath", "error", err) |
| 128 | h.d.Render.HTTPError(w, r, http.StatusInternalServerError, "") |
| 129 | return |
| 130 | } |
| 131 | if kind == repogit.EntryBlob { |
| 132 | http.Redirect(w, r, "/"+cc.owner+"/"+cc.row.Name+"/blob/"+cc.ref+"/"+cc.subpath, http.StatusSeeOther) |
| 133 | return |
| 134 | } |
| 135 | entries, err := repogit.LsTree(r.Context(), cc.gitDir, cc.ref, cc.subpath) |
| 136 | if err != nil { |
| 137 | if errors.Is(err, repogit.ErrNotATree) { |
| 138 | h.d.Render.HTTPError(w, r, http.StatusNotFound, "") |
| 139 | return |
| 140 | } |
| 141 | h.d.Logger.WarnContext(r.Context(), "code: LsTree", "error", err) |
| 142 | h.d.Render.HTTPError(w, r, http.StatusInternalServerError, "") |
| 143 | return |
| 144 | } |
| 145 | // README detection on the requested directory only. |
| 146 | readmeHTML := h.findAndRenderREADME(r, cc, entries) |
| 147 | head, headFound, headErr := repogit.HeadOf(r.Context(), cc.gitDir, cc.ref) |
| 148 | if headErr != nil { |
| 149 | h.d.Logger.WarnContext(r.Context(), "code: HeadOf", "error", headErr) |
| 150 | } |
| 151 | topics, _ := h.rq.ListRepoTopics(r.Context(), h.d.Pool, cc.row.ID) |
| 152 | |
| 153 | h.d.Render.RenderPage(w, r, "repo/tree", map[string]any{ |
| 154 | "Title": cc.row.Name + " · " + cc.owner, |
| 155 | "CSRFToken": middleware.CSRFTokenForRequest(r), |
| 156 | "Owner": cc.owner, |
| 157 | "Repo": cc.row, |
| 158 | "Ref": cc.ref, |
| 159 | "Path": cc.subpath, |
| 160 | "Crumbs": breadcrumbs(cc.owner, cc.row.Name, cc.ref, cc.subpath), |
| 161 | "Entries": entries, |
| 162 | "Branches": cc.refs.Branches, |
| 163 | "Tags": cc.refs.Tags, |
| 164 | "Head": head, |
| 165 | "HeadFound": headFound, |
| 166 | "README": template.HTML(readmeHTML), //nolint:gosec // sanitized by mdrender |
| 167 | "HTTPSCloneURL": h.cloneHTTPS(cc.owner, cc.row.Name), |
| 168 | "SSHEnabled": h.d.CloneURLs.SSHEnabled, |
| 169 | "SSHCloneURL": h.cloneSSH(cc.owner, cc.row.Name), |
| 170 | "RepoTopics": topics, |
| 171 | "RepoCounts": h.subnavCounts(r.Context(), cc.row.ID, cc.row.ForkCount), |
| 172 | "CanSettings": h.canViewSettings(middleware.CurrentUserFromContext(r.Context())), |
| 173 | "ActiveSubnav": "code", |
| 174 | }) |
| 175 | } |
| 176 | |
| 177 | func refNames(refs repogit.RefListing) []string { |
| 178 | allNames := make([]string, 0, len(refs.Branches)+len(refs.Tags)) |
| 179 | for _, b := range refs.Branches { |
| 180 | allNames = append(allNames, b.Name) |
| 181 | } |
| 182 | for _, t := range refs.Tags { |
| 183 | allNames = append(allNames, t.Name) |
| 184 | } |
| 185 | return allNames |
| 186 | } |
| 187 | |
| 188 | // findAndRenderREADME looks for README* in the supplied entries (case- |
| 189 | // insensitive). Returns rendered HTML for markdown sources; returns a |
| 190 | // `<pre>`-wrapped escaped string for non-markdown text. Empty when |
| 191 | // no README is present. |
| 192 | func (h *Handlers) findAndRenderREADME(r *http.Request, cc *codeContext, entries []repogit.TreeEntry) string { |
| 193 | const maxREADMEBytes = 1 * 1024 * 1024 // 1 MiB cap |
| 194 | for _, e := range entries { |
| 195 | if e.Kind != repogit.EntryBlob { |
| 196 | continue |
| 197 | } |
| 198 | lower := strings.ToLower(e.Name) |
| 199 | if !strings.HasPrefix(lower, "readme") { |
| 200 | continue |
| 201 | } |
| 202 | full := joinPath(cc.subpath, e.Name) |
| 203 | body, err := repogit.ReadBlobBytes(r.Context(), cc.gitDir, cc.ref, full, maxREADMEBytes) |
| 204 | if err != nil && !errors.Is(err, repogit.ErrBlobTooLarge) { |
| 205 | return "" |
| 206 | } |
| 207 | // Markdown: render via Goldmark + sanitizer. |
| 208 | if hasExt(lower, []string{".md", ".markdown"}) { |
| 209 | out, mderr := mdrender.RenderHTML(body) |
| 210 | if mderr == nil { |
| 211 | return out |
| 212 | } |
| 213 | } |
| 214 | // Non-markdown plain text: escape + <pre>. |
| 215 | return "<pre class=\"shithub-readme-plain\">" + template.HTMLEscapeString(string(body)) + "</pre>" |
| 216 | } |
| 217 | return "" |
| 218 | } |
| 219 | |
| 220 | func hasExt(filename string, exts []string) bool { |
| 221 | for _, e := range exts { |
| 222 | if strings.HasSuffix(filename, e) { |
| 223 | return true |
| 224 | } |
| 225 | } |
| 226 | return false |
| 227 | } |
| 228 | |
| 229 | // codeBlob renders the file viewer. |
| 230 | func (h *Handlers) codeBlob(w http.ResponseWriter, r *http.Request) { |
| 231 | cc, ok := h.loadCodeContext(w, r) |
| 232 | if !ok { |
| 233 | return |
| 234 | } |
| 235 | kind, _, size, err := repogit.StatPath(r.Context(), cc.gitDir, cc.ref, cc.subpath) |
| 236 | if err != nil || kind != repogit.EntryBlob { |
| 237 | h.d.Render.HTTPError(w, r, http.StatusNotFound, "") |
| 238 | return |
| 239 | } |
| 240 | const largeFileThreshold = 1 * 1024 * 1024 // 1 MiB |
| 241 | const maxReadBytes = 4 * 1024 * 1024 // never read more than 4 MiB even for highlighting |
| 242 | |
| 243 | data := map[string]any{ |
| 244 | "Title": cc.subpath + " · " + cc.row.Name, |
| 245 | "CSRFToken": middleware.CSRFTokenForRequest(r), |
| 246 | "Owner": cc.owner, |
| 247 | "Repo": cc.row, |
| 248 | "Ref": cc.ref, |
| 249 | "Path": cc.subpath, |
| 250 | "Crumbs": breadcrumbs(cc.owner, cc.row.Name, cc.ref, cc.subpath), |
| 251 | "Branches": cc.refs.Branches, |
| 252 | "Tags": cc.refs.Tags, |
| 253 | "Size": size, |
| 254 | "IsLarge": size > largeFileThreshold, |
| 255 | "IsBinary": false, |
| 256 | "IsImage": false, |
| 257 | "IsMarkdown": false, |
| 258 | "Language": highlight.LanguageGuess(cc.subpath), |
| 259 | "RepoCounts": h.subnavCounts(r.Context(), cc.row.ID, cc.row.ForkCount), |
| 260 | "CanSettings": h.canViewSettings(middleware.CurrentUserFromContext(r.Context())), |
| 261 | "ActiveSubnav": "code", |
| 262 | } |
| 263 | if size > largeFileThreshold { |
| 264 | h.d.Render.RenderPage(w, r, "repo/blob", data) |
| 265 | return |
| 266 | } |
| 267 | body, err := repogit.ReadBlobBytes(r.Context(), cc.gitDir, cc.ref, cc.subpath, maxReadBytes) |
| 268 | if err != nil && !errors.Is(err, repogit.ErrBlobTooLarge) { |
| 269 | h.d.Render.HTTPError(w, r, http.StatusInternalServerError, "") |
| 270 | return |
| 271 | } |
| 272 | if isBinary(body) { |
| 273 | data["IsBinary"] = true |
| 274 | if isImageExt(cc.subpath) && size <= 5*1024*1024 { |
| 275 | data["IsImage"] = true |
| 276 | } |
| 277 | h.d.Render.RenderPage(w, r, "repo/blob", data) |
| 278 | return |
| 279 | } |
| 280 | // Text path: highlight or markdown-render. |
| 281 | if hasExt(strings.ToLower(cc.subpath), []string{".md", ".markdown"}) { |
| 282 | data["IsMarkdown"] = true |
| 283 | rendered, mderr := mdrender.RenderHTML(body) |
| 284 | if mderr == nil { |
| 285 | data["MarkdownHTML"] = template.HTML(rendered) //nolint:gosec // sanitized |
| 286 | } |
| 287 | data["RawSource"] = string(body) |
| 288 | } |
| 289 | // Per-line highlighted fragments. The template composes the row |
| 290 | // table; chroma only colors the tokens inside each line. |
| 291 | data["Lines"] = highlight.RenderLines(cc.subpath, string(body)) |
| 292 | h.d.Render.RenderPage(w, r, "repo/blob", data) |
| 293 | } |
| 294 | |
| 295 | // codeRaw streams the raw bytes. Force `attachment` for executable |
| 296 | // content types (HTML/SVG/JS/etc.) since shithub doesn't have a |
| 297 | // separate raw host. |
| 298 | func (h *Handlers) codeRaw(w http.ResponseWriter, r *http.Request) { |
| 299 | cc, ok := h.loadCodeContext(w, r) |
| 300 | if !ok { |
| 301 | return |
| 302 | } |
| 303 | kind, _, size, err := repogit.StatPath(r.Context(), cc.gitDir, cc.ref, cc.subpath) |
| 304 | if err != nil || kind != repogit.EntryBlob { |
| 305 | h.d.Render.HTTPError(w, r, http.StatusNotFound, "") |
| 306 | return |
| 307 | } |
| 308 | contentType, forceAttachment := rawContentType(cc.subpath) |
| 309 | w.Header().Set("Content-Type", contentType) |
| 310 | w.Header().Set("Cache-Control", "private, max-age=0, must-revalidate") |
| 311 | w.Header().Set("X-Content-Type-Options", "nosniff") |
| 312 | w.Header().Set("Content-Security-Policy", "default-src 'none'; sandbox") |
| 313 | if forceAttachment { |
| 314 | w.Header().Set("Content-Disposition", `attachment; filename="`+path.Base(cc.subpath)+`"`) |
| 315 | } |
| 316 | if size > 0 { |
| 317 | w.Header().Set("Content-Length", fmt.Sprintf("%d", size)) |
| 318 | } |
| 319 | if err := repogit.StreamBlob(r.Context(), cc.gitDir, cc.ref, cc.subpath, w); err != nil { |
| 320 | h.d.Logger.WarnContext(r.Context(), "code: stream raw", "error", err) |
| 321 | } |
| 322 | } |
| 323 | |
| 324 | // codeFinder serves /find/{ref} — full list pre-filtered by `q`. |
| 325 | func (h *Handlers) codeFinder(w http.ResponseWriter, r *http.Request) { |
| 326 | cc, ok := h.loadCodeContext(w, r) |
| 327 | if !ok { |
| 328 | return |
| 329 | } |
| 330 | paths, err := repogit.ListAllPaths(r.Context(), cc.gitDir, cc.ref) |
| 331 | if err != nil { |
| 332 | h.d.Logger.WarnContext(r.Context(), "code: ListAllPaths", "error", err) |
| 333 | h.d.Render.HTTPError(w, r, http.StatusInternalServerError, "") |
| 334 | return |
| 335 | } |
| 336 | q := r.URL.Query().Get("q") |
| 337 | matches := finder.Filter(paths, q, 200) |
| 338 | h.d.Render.RenderPage(w, r, "repo/finder", map[string]any{ |
| 339 | "Title": "Find file · " + cc.row.Name, |
| 340 | "CSRFToken": middleware.CSRFTokenForRequest(r), |
| 341 | "Owner": cc.owner, |
| 342 | "Repo": cc.row, |
| 343 | "Ref": cc.ref, |
| 344 | "Query": q, |
| 345 | "Matches": matches, |
| 346 | "Branches": cc.refs.Branches, |
| 347 | "Tags": cc.refs.Tags, |
| 348 | }) |
| 349 | } |
| 350 | |
| 351 | // breadcrumbs returns the click-each-segment slice for the tree/blob |
| 352 | // header. |
| 353 | type Breadcrumb struct { |
| 354 | Name string |
| 355 | URL string |
| 356 | } |
| 357 | |
| 358 | func breadcrumbs(owner, repoName, ref, subpath string) []Breadcrumb { |
| 359 | out := []Breadcrumb{ |
| 360 | {Name: repoName, URL: fmt.Sprintf("/%s/%s/tree/%s", owner, repoName, ref)}, |
| 361 | } |
| 362 | if subpath == "" { |
| 363 | return out |
| 364 | } |
| 365 | parts := strings.Split(subpath, "/") |
| 366 | for i, p := range parts { |
| 367 | out = append(out, Breadcrumb{ |
| 368 | Name: p, |
| 369 | URL: fmt.Sprintf("/%s/%s/tree/%s/%s", owner, repoName, ref, strings.Join(parts[:i+1], "/")), |
| 370 | }) |
| 371 | } |
| 372 | return out |
| 373 | } |
| 374 | |
| 375 | // validateSubpath is the path-traversal guard. Reject `..`, control |
| 376 | // chars, leading slash, and `\`. |
| 377 | func validateSubpath(p string) bool { |
| 378 | if p == "" { |
| 379 | return true |
| 380 | } |
| 381 | if strings.HasPrefix(p, "/") || strings.Contains(p, "\\") { |
| 382 | return false |
| 383 | } |
| 384 | for _, seg := range strings.Split(p, "/") { |
| 385 | if seg == "" || seg == ".." { |
| 386 | return false |
| 387 | } |
| 388 | for _, c := range seg { |
| 389 | if c < 0x20 || c == 0x7f { |
| 390 | return false |
| 391 | } |
| 392 | } |
| 393 | } |
| 394 | return true |
| 395 | } |
| 396 | |
| 397 | func isHex(s string) bool { |
| 398 | for _, c := range s { |
| 399 | switch { |
| 400 | case c >= '0' && c <= '9', c >= 'a' && c <= 'f', c >= 'A' && c <= 'F': |
| 401 | default: |
| 402 | return false |
| 403 | } |
| 404 | } |
| 405 | return true |
| 406 | } |
| 407 | |
| 408 | // rawContentType maps an extension to (Content-Type, forceAttachment). |
| 409 | // Executable content types force `attachment` to defeat XSS via raw view |
| 410 | // (no separate raw.host yet). |
| 411 | func rawContentType(p string) (string, bool) { |
| 412 | ext := strings.ToLower(path.Ext(p)) |
| 413 | switch ext { |
| 414 | case ".html", ".htm", ".xhtml", ".svg", ".js", ".mjs", ".wasm": |
| 415 | return "text/plain; charset=utf-8", true |
| 416 | case ".png": |
| 417 | return "image/png", false |
| 418 | case ".jpg", ".jpeg": |
| 419 | return "image/jpeg", false |
| 420 | case ".gif": |
| 421 | return "image/gif", false |
| 422 | case ".webp": |
| 423 | return "image/webp", false |
| 424 | case ".pdf": |
| 425 | return "application/pdf", false |
| 426 | case ".css": |
| 427 | return "text/css; charset=utf-8", false |
| 428 | case ".json": |
| 429 | return "application/json; charset=utf-8", false |
| 430 | case ".txt", ".md", ".markdown", ".yml", ".yaml", ".toml", ".log": |
| 431 | return "text/plain; charset=utf-8", false |
| 432 | default: |
| 433 | // Sniff by inspecting the body would be ideal, but we already |
| 434 | // stream — fall back to text/plain for safety. |
| 435 | return "text/plain; charset=utf-8", false |
| 436 | } |
| 437 | } |
| 438 | |
| 439 | func isImageExt(p string) bool { |
| 440 | switch strings.ToLower(path.Ext(p)) { |
| 441 | case ".png", ".jpg", ".jpeg", ".gif", ".webp": |
| 442 | return true |
| 443 | } |
| 444 | return false |
| 445 | } |
| 446 | |
| 447 | // isBinary scans the first 8 KiB for a NUL byte. |
| 448 | func isBinary(b []byte) bool { |
| 449 | const window = 8192 |
| 450 | if len(b) > window { |
| 451 | b = b[:window] |
| 452 | } |
| 453 | return bytes.IndexByte(b, 0) >= 0 |
| 454 | } |
| 455 | |
| 456 | // joinPath joins two slash-separated paths, ignoring an empty parent. |
| 457 | func joinPath(parent, child string) string { |
| 458 | if parent == "" { |
| 459 | return child |
| 460 | } |
| 461 | return parent + "/" + child |
| 462 | } |
| 463 | |
| 464 | // silence pgtype unused-import warning when the loadRepoAndAuthorize |
| 465 | // helper is in this file's package but defined elsewhere. |
| 466 | var _ = pgtype.Int8{} |
| 467 |