diff --git a/go/vcs/discovery.go b/go/vcs/discovery.go new file mode 100644 index 00000000..7d614598 --- /dev/null +++ b/go/vcs/discovery.go @@ -0,0 +1,55 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package vcs + +import ( + "encoding/xml" + "io" + "strings" +) + +// parseMetaGoImports returns meta imports from the HTML in r. +// Parsing ends at the end of the
section or the beginning of the . +func parseMetaGoImports(r io.Reader) (imports []metaImport) { + d := xml.NewDecoder(r) + d.Strict = false + for { + t, err := d.Token() + if err != nil { + return + } + if e, ok := t.(xml.StartElement); ok && strings.EqualFold(e.Name.Local, "body") { + return + } + if e, ok := t.(xml.EndElement); ok && strings.EqualFold(e.Name.Local, "head") { + return + } + e, ok := t.(xml.StartElement) + if !ok || !strings.EqualFold(e.Name.Local, "meta") { + continue + } + if attrValue(e.Attr, "name") != "go-import" { + continue + } + if f := strings.Fields(attrValue(e.Attr, "content")); len(f) == 3 { + imports = append(imports, metaImport{ + Prefix: f[0], + VCS: f[1], + RepoRoot: f[2], + }) + } + } +} + +// attrValue returns the attribute value for the case-insensitive key +// `name', or the empty string if nothing is found. +func attrValue(attrs []xml.Attr, name string) string { + for _, a := range attrs { + if strings.EqualFold(a.Name.Local, name) { + return a.Value + } + } + return "" +} diff --git a/go/vcs/env.go b/go/vcs/env.go new file mode 100644 index 00000000..e846f5b3 --- /dev/null +++ b/go/vcs/env.go @@ -0,0 +1,39 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package vcs + +import ( + "os" + "strings" +) + +// envForDir returns a copy of the environment +// suitable for running in the given directory. +// The environment is the current process's environment +// but with an updated $PWD, so that an os.Getwd in the +// child will be faster. +func envForDir(dir string) []string { + env := os.Environ() + // Internally we only use rooted paths, so dir is rooted. + // Even if dir is not rooted, no harm done. + return mergeEnvLists([]string{"PWD=" + dir}, env) +} + +// mergeEnvLists merges the two environment lists such that +// variables with the same name in "in" replace those in "out". +func mergeEnvLists(in, out []string) []string { +NextVar: + for _, inkv := range in { + k := strings.SplitAfterN(inkv, "=", 2)[0] + for i, outkv := range out { + if strings.HasPrefix(outkv, k) { + out[i] = inkv + continue NextVar + } + } + out = append(out, inkv) + } + return out +} diff --git a/go/vcs/http.go b/go/vcs/http.go new file mode 100644 index 00000000..96188185 --- /dev/null +++ b/go/vcs/http.go @@ -0,0 +1,80 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package vcs + +import ( + "fmt" + "io" + "io/ioutil" + "log" + "net/http" + "net/url" +) + +// httpClient is the default HTTP client, but a variable so it can be +// changed by tests, without modifying http.DefaultClient. +var httpClient = http.DefaultClient + +// httpGET returns the data from an HTTP GET request for the given URL. +func httpGET(url string) ([]byte, error) { + resp, err := httpClient.Get(url) + if err != nil { + return nil, err + } + defer resp.Body.Close() + if resp.StatusCode != 200 { + return nil, fmt.Errorf("%s: %s", url, resp.Status) + } + b, err := ioutil.ReadAll(resp.Body) + if err != nil { + return nil, fmt.Errorf("%s: %v", url, err) + } + return b, nil +} + +// httpsOrHTTP returns the body of either the importPath's +// https resource or, if unavailable, the http resource. +func httpsOrHTTP(importPath string) (urlStr string, body io.ReadCloser, err error) { + fetch := func(scheme string) (urlStr string, res *http.Response, err error) { + u, err := url.Parse(scheme + "://" + importPath) + if err != nil { + return "", nil, err + } + u.RawQuery = "go-get=1" + urlStr = u.String() + if Verbose { + log.Printf("Fetching %s", urlStr) + } + res, err = httpClient.Get(urlStr) + return + } + closeBody := func(res *http.Response) { + if res != nil { + res.Body.Close() + } + } + urlStr, res, err := fetch("https") + if err != nil || res.StatusCode != 200 { + if Verbose { + if err != nil { + log.Printf("https fetch failed.") + } else { + log.Printf("ignoring https fetch with status code %d", res.StatusCode) + } + } + closeBody(res) + urlStr, res, err = fetch("http") + } + if err != nil { + closeBody(res) + return "", nil, err + } + // Note: accepting a non-200 OK here, so people can serve a + // meta import in their http 404 page. + if Verbose { + log.Printf("Parsing meta tags from %s (status code %d)", urlStr, res.StatusCode) + } + return urlStr, res.Body, nil +} diff --git a/go/vcs/vcs.go b/go/vcs/vcs.go new file mode 100644 index 00000000..4b27e163 --- /dev/null +++ b/go/vcs/vcs.go @@ -0,0 +1,697 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package vcs + +import ( + "bytes" + "encoding/json" + "errors" + "fmt" + "log" + "os" + "os/exec" + "path/filepath" + "regexp" + "strings" +) + +// Verbose enables verbose operation logging. +var Verbose bool + +// ShowCmd controls whether VCS commands are printed. +var ShowCmd bool + +// A Cmd describes how to use a version control system +// like Mercurial, Git, or Subversion. +type Cmd struct { + Name string + Cmd string // name of binary to invoke command + + CreateCmd string // command to download a fresh copy of a repository + DownloadCmd string // command to download updates into an existing repository + + TagCmd []TagCmd // commands to list tags + TagLookupCmd []TagCmd // commands to lookup tags before running tagSyncCmd + TagSyncCmd string // command to sync to specific tag + TagSyncDefault string // command to sync to default tag + + Scheme []string + PingCmd string +} + +// A TagCmd describes a command to list available tags +// that can be passed to Cmd.TagSyncCmd. +type TagCmd struct { + Cmd string // command to list tags + Pattern string // regexp to extract tags from list +} + +// vcsList lists the known version control systems +var vcsList = []*Cmd{ + vcsHg, + vcsGit, + vcsSvn, + vcsBzr, +} + +// ByCmd returns the version control system for the given +// command name (hg, git, svn, bzr). +func ByCmd(cmd string) *Cmd { + for _, vcs := range vcsList { + if vcs.Cmd == cmd { + return vcs + } + } + return nil +} + +// vcsHg describes how to use Mercurial. +var vcsHg = &Cmd{ + Name: "Mercurial", + Cmd: "hg", + + CreateCmd: "clone -U {repo} {dir}", + DownloadCmd: "pull", + + // We allow both tag and branch names as 'tags' + // for selecting a version. This lets people have + // a go.release.r60 branch and a go1 branch + // and make changes in both, without constantly + // editing .hgtags. + TagCmd: []TagCmd{ + {"tags", `^(\S+)`}, + {"branches", `^(\S+)`}, + }, + TagSyncCmd: "update -r {tag}", + TagSyncDefault: "update default", + + Scheme: []string{"https", "http", "ssh"}, + PingCmd: "identify {scheme}://{repo}", +} + +// vcsGit describes how to use Git. +var vcsGit = &Cmd{ + Name: "Git", + Cmd: "git", + + CreateCmd: "clone {repo} {dir}", + DownloadCmd: "pull --ff-only", + + TagCmd: []TagCmd{ + // tags/xxx matches a git tag named xxx + // origin/xxx matches a git branch named xxx on the default remote repository + {"show-ref", `(?:tags|origin)/(\S+)$`}, + }, + TagLookupCmd: []TagCmd{ + {"show-ref tags/{tag} origin/{tag}", `((?:tags|origin)/\S+)$`}, + }, + TagSyncCmd: "checkout {tag}", + TagSyncDefault: "checkout master", + + Scheme: []string{"git", "https", "http", "git+ssh"}, + PingCmd: "ls-remote {scheme}://{repo}", +} + +// vcsBzr describes how to use Bazaar. +var vcsBzr = &Cmd{ + Name: "Bazaar", + Cmd: "bzr", + + CreateCmd: "branch {repo} {dir}", + + // Without --overwrite bzr will not pull tags that changed. + // Replace by --overwrite-tags after http://pad.lv/681792 goes in. + DownloadCmd: "pull --overwrite", + + TagCmd: []TagCmd{{"tags", `^(\S+)`}}, + TagSyncCmd: "update -r {tag}", + TagSyncDefault: "update -r revno:-1", + + Scheme: []string{"https", "http", "bzr", "bzr+ssh"}, + PingCmd: "info {scheme}://{repo}", +} + +// vcsSvn describes how to use Subversion. +var vcsSvn = &Cmd{ + Name: "Subversion", + Cmd: "svn", + + CreateCmd: "checkout {repo} {dir}", + DownloadCmd: "update", + + // There is no tag command in subversion. + // The branch information is all in the path names. + + Scheme: []string{"https", "http", "svn", "svn+ssh"}, + PingCmd: "info {scheme}://{repo}", +} + +func (v *Cmd) String() string { + return v.Name +} + +// run runs the command line cmd in the given directory. +// keyval is a list of key, value pairs. run expands +// instances of {key} in cmd into value, but only after +// splitting cmd into individual arguments. +// If an error occurs, run prints the command line and the +// command's combined stdout+stderr to standard error. +// Otherwise run discards the command's output. +func (v *Cmd) run(dir string, cmd string, keyval ...string) error { + _, err := v.run1(dir, cmd, keyval, true) + return err +} + +// runVerboseOnly is like run but only generates error output to standard error in verbose mode. +func (v *Cmd) runVerboseOnly(dir string, cmd string, keyval ...string) error { + _, err := v.run1(dir, cmd, keyval, false) + return err +} + +// runOutput is like run but returns the output of the command. +func (v *Cmd) runOutput(dir string, cmd string, keyval ...string) ([]byte, error) { + return v.run1(dir, cmd, keyval, true) +} + +// run1 is the generalized implementation of run and runOutput. +func (v *Cmd) run1(dir string, cmdline string, keyval []string, verbose bool) ([]byte, error) { + m := make(map[string]string) + for i := 0; i < len(keyval); i += 2 { + m[keyval[i]] = keyval[i+1] + } + args := strings.Fields(cmdline) + for i, arg := range args { + args[i] = expand(m, arg) + } + + _, err := exec.LookPath(v.Cmd) + if err != nil { + fmt.Fprintf(os.Stderr, + "go: missing %s command. See http://golang.org/s/gogetcmd\n", + v.Name) + return nil, err + } + + cmd := exec.Command(v.Cmd, args...) + cmd.Dir = dir + cmd.Env = envForDir(cmd.Dir) + if ShowCmd { + fmt.Printf("cd %s\n", dir) + fmt.Printf("%s %s\n", v.Cmd, strings.Join(args, " ")) + } + var buf bytes.Buffer + cmd.Stdout = &buf + cmd.Stderr = &buf + err = cmd.Run() + out := buf.Bytes() + if err != nil { + if verbose || Verbose { + fmt.Fprintf(os.Stderr, "# cd %s; %s %s\n", dir, v.Cmd, strings.Join(args, " ")) + os.Stderr.Write(out) + } + return nil, err + } + return out, nil +} + +// Ping pings the repo to determine if scheme used if valid. +// This repo must be pingable with this scheme and VCS. +func (v *Cmd) Ping(scheme, repo string) error { + return v.runVerboseOnly(".", v.PingCmd, "scheme", scheme, "repo", repo) +} + +// Create creates a new copy of repo in dir. +// The parent of dir must exist; dir must not. +func (v *Cmd) Create(dir, repo string) error { + return v.run(".", v.CreateCmd, "dir", dir, "repo", repo) +} + +// Download downloads any new changes for the repo in dir. +// dir must be a valid VCS repo compatible with v. +func (v *Cmd) Download(dir string) error { + return v.run(dir, v.DownloadCmd) +} + +// Tags returns the list of available tags for the repo in dir. +// dir must be a valid VCS repo compatible with v. +func (v *Cmd) Tags(dir string) ([]string, error) { + var tags []string + for _, tc := range v.TagCmd { + out, err := v.runOutput(dir, tc.Cmd) + if err != nil { + return nil, err + } + re := regexp.MustCompile(`(?m-s)` + tc.Pattern) + for _, m := range re.FindAllStringSubmatch(string(out), -1) { + tags = append(tags, m[1]) + } + } + return tags, nil +} + +// TagSync syncs the repo in dir to the named tag, +// which either is a tag returned by tags or is v.TagDefault. +// dir must be a valid VCS repo compatible with v and the tag must exist. +func (v *Cmd) TagSync(dir, tag string) error { + if v.TagSyncCmd == "" { + return nil + } + if tag != "" { + for _, tc := range v.TagLookupCmd { + out, err := v.runOutput(dir, tc.Cmd, "tag", tag) + if err != nil { + return err + } + re := regexp.MustCompile(`(?m-s)` + tc.Pattern) + m := re.FindStringSubmatch(string(out)) + if len(m) > 1 { + tag = m[1] + break + } + } + } + if tag == "" && v.TagSyncDefault != "" { + return v.run(dir, v.TagSyncDefault) + } + return v.run(dir, v.TagSyncCmd, "tag", tag) +} + +// A vcsPath describes how to convert an import path into a +// version control system and repository name. +type vcsPath struct { + prefix string // prefix this description applies to + re string // pattern for import path + repo string // repository to use (expand with match of re) + vcs string // version control system to use (expand with match of re) + check func(match map[string]string) error // additional checks + ping bool // ping for scheme to use to download repo + + regexp *regexp.Regexp // cached compiled form of re +} + +// FromDir inspects dir and its parents to determine the +// version control system and code repository to use. +// On return, root is the import path +// corresponding to the root of the repository +// (thus root is a prefix of importPath). +func FromDir(dir, srcRoot string) (vcs *Cmd, root string, err error) { + // Clean and double-check that dir is in (a subdirectory of) srcRoot. + dir = filepath.Clean(dir) + srcRoot = filepath.Clean(srcRoot) + if len(dir) <= len(srcRoot) || dir[len(srcRoot)] != filepath.Separator { + return nil, "", fmt.Errorf("directory %q is outside source root %q", dir, srcRoot) + } + + for len(dir) > len(srcRoot) { + for _, vcs := range vcsList { + if fi, err := os.Stat(filepath.Join(dir, "."+vcs.Cmd)); err == nil && fi.IsDir() { + return vcs, dir[len(srcRoot)+1:], nil + } + } + + // Move to parent. + ndir := filepath.Dir(dir) + if len(ndir) >= len(dir) { + // Shouldn't happen, but just in case, stop. + break + } + dir = ndir + } + + return nil, "", fmt.Errorf("directory %q is not using a known version control system", dir) +} + +// RepoRoot represents a version control system, a repo, and a root of +// where to put it on disk. +type RepoRoot struct { + VCS *Cmd + + // repo is the repository URL, including scheme + Repo string + + // root is the import path corresponding to the root of the + // repository + Root string +} + +// RepoRootForImportPath analyzes importPath to determine the +// version control system, and code repository to use. +func RepoRootForImportPath(importPath string, verbose bool) (*RepoRoot, error) { + rr, err := RepoRootForImportPathStatic(importPath, "") + if err == errUnknownSite { + rr, err = RepoRootForImportDynamic(importPath, verbose) + + // RepoRootForImportDynamic returns error detail + // that is irrelevant if the user didn't intend to use a + // dynamic import in the first place. + // Squelch it. + if err != nil { + if Verbose { + log.Printf("import %q: %v", importPath, err) + } + err = fmt.Errorf("unrecognized import path %q", importPath) + } + } + + if err == nil && strings.Contains(importPath, "...") && strings.Contains(rr.Root, "...") { + // Do not allow wildcards in the repo root. + rr = nil + err = fmt.Errorf("cannot expand ... in %q", importPath) + } + return rr, err +} + +var errUnknownSite = errors.New("dynamic lookup required to find mapping") + +// RepoRootForImportPathStatic attempts to map importPath to a +// RepoRoot using the commonly-used VCS hosting sites in vcsPaths +// (github.com/user/dir), or from a fully-qualified importPath already +// containing its VCS type (foo.com/repo.git/dir) +// +// If scheme is non-empty, that scheme is forced. +func RepoRootForImportPathStatic(importPath, scheme string) (*RepoRoot, error) { + if strings.Contains(importPath, "://") { + return nil, fmt.Errorf("invalid import path %q", importPath) + } + for _, srv := range vcsPaths { + if !strings.HasPrefix(importPath, srv.prefix) { + continue + } + m := srv.regexp.FindStringSubmatch(importPath) + if m == nil { + if srv.prefix != "" { + return nil, fmt.Errorf("invalid %s import path %q", srv.prefix, importPath) + } + continue + } + + // Build map of named subexpression matches for expand. + match := map[string]string{ + "prefix": srv.prefix, + "import": importPath, + } + for i, name := range srv.regexp.SubexpNames() { + if name != "" && match[name] == "" { + match[name] = m[i] + } + } + if srv.vcs != "" { + match["vcs"] = expand(match, srv.vcs) + } + if srv.repo != "" { + match["repo"] = expand(match, srv.repo) + } + if srv.check != nil { + if err := srv.check(match); err != nil { + return nil, err + } + } + vcs := ByCmd(match["vcs"]) + if vcs == nil { + return nil, fmt.Errorf("unknown version control system %q", match["vcs"]) + } + if srv.ping { + if scheme != "" { + match["repo"] = scheme + "://" + match["repo"] + } else { + for _, scheme := range vcs.Scheme { + if vcs.Ping(scheme, match["repo"]) == nil { + match["repo"] = scheme + "://" + match["repo"] + break + } + } + } + } + rr := &RepoRoot{ + VCS: vcs, + Repo: match["repo"], + Root: match["root"], + } + return rr, nil + } + return nil, errUnknownSite +} + +// RepoRootForImportDynamic finds a *RepoRoot for a custom domain that's not +// statically known by RepoRootForImportPathStatic. +// +// This handles "vanity import paths" like "name.tld/pkg/foo". +func RepoRootForImportDynamic(importPath string, verbose bool) (*RepoRoot, error) { + slash := strings.Index(importPath, "/") + if slash < 0 { + return nil, errors.New("import path doesn't contain a slash") + } + host := importPath[:slash] + if !strings.Contains(host, ".") { + return nil, errors.New("import path doesn't contain a hostname") + } + urlStr, body, err := httpsOrHTTP(importPath) + if err != nil { + return nil, fmt.Errorf("http/https fetch: %v", err) + } + defer body.Close() + metaImport, err := matchGoImport(parseMetaGoImports(body), importPath) + if err != nil { + if err != errNoMatch { + return nil, fmt.Errorf("parse %s: %v", urlStr, err) + } + return nil, fmt.Errorf("parse %s: no go-import meta tags", urlStr) + } + if verbose { + log.Printf("get %q: found meta tag %#v at %s", importPath, metaImport, urlStr) + } + // If the import was "uni.edu/bob/project", which said the + // prefix was "uni.edu" and the RepoRoot was "evilroot.com", + // make sure we don't trust Bob and check out evilroot.com to + // "uni.edu" yet (possibly overwriting/preempting another + // non-evil student). Instead, first verify the root and see + // if it matches Bob's claim. + if metaImport.Prefix != importPath { + if verbose { + log.Printf("get %q: verifying non-authoritative meta tag", importPath) + } + urlStr0 := urlStr + urlStr, body, err = httpsOrHTTP(metaImport.Prefix) + if err != nil { + return nil, fmt.Errorf("fetch %s: %v", urlStr, err) + } + imports := parseMetaGoImports(body) + if len(imports) == 0 { + return nil, fmt.Errorf("fetch %s: no go-import meta tag", urlStr) + } + metaImport2, err := matchGoImport(imports, importPath) + if err != nil || metaImport != metaImport2 { + return nil, fmt.Errorf("%s and %s disagree about go-import for %s", urlStr0, urlStr, metaImport.Prefix) + } + } + + if !strings.Contains(metaImport.RepoRoot, "://") { + return nil, fmt.Errorf("%s: invalid repo root %q; no scheme", urlStr, metaImport.RepoRoot) + } + rr := &RepoRoot{ + VCS: ByCmd(metaImport.VCS), + Repo: metaImport.RepoRoot, + Root: metaImport.Prefix, + } + if rr.VCS == nil { + return nil, fmt.Errorf("%s: unknown vcs %q", urlStr, metaImport.VCS) + } + return rr, nil +} + +// metaImport represents the parsed tags from HTML files. +type metaImport struct { + Prefix, VCS, RepoRoot string +} + +// errNoMatch is returned from matchGoImport when there's no applicable match. +var errNoMatch = errors.New("no import match") + +// matchGoImport returns the metaImport from imports matching importPath. +// An error is returned if there are multiple matches. +// errNoMatch is returned if none match. +func matchGoImport(imports []metaImport, importPath string) (_ metaImport, err error) { + match := -1 + for i, im := range imports { + if !strings.HasPrefix(importPath, im.Prefix) { + continue + } + if match != -1 { + err = fmt.Errorf("multiple meta tags match import path %q", importPath) + return + } + match = i + } + if match == -1 { + err = errNoMatch + return + } + return imports[match], nil +} + +// expand rewrites s to replace {k} with match[k] for each key k in match. +func expand(match map[string]string, s string) string { + for k, v := range match { + s = strings.Replace(s, "{"+k+"}", v, -1) + } + return s +} + +// vcsPaths lists the known vcs paths. +var vcsPaths = []*vcsPath{ + // Google Code - new syntax + { + prefix: "code.google.com/", + re: `^(?P