Move Dockerfile parsing to a dedicated package, add support for RUN --mount=type=bind,from=...

.github/workflows/ci.yml

@@ -55,6 +55,14 @@ jobs:
       #    files: coverage.out
       #    fail_ci_if_error: true
       #    verbose: true
+      # in the meantime, upload coverage to GHA
+      - run: docker run --rm -i test go tool cover -html /dev/stdin -o /dev/stdout < coverage.out > coverage.html
+      - uses: actions/upload-artifact@v4
+        with:
+          name: coverage
+          path: coverage.*
+          include-hidden-files: true
+          if-no-files-found: error
   dockerfile:
     name: Test Dockerfile
     runs-on: ubuntu-latest

Dockerfile

@@ -1,4 +1,4 @@
-FROM golang:1.20-bullseye AS build
+FROM golang:1.21-bookworm AS build
 
 SHELL ["bash", "-Eeuo", "pipefail", "-xc"]
 

Dockerfile.release

@@ -1,4 +1,4 @@
-FROM golang:1.20-bullseye
+FROM golang:1.21-bookworm
 
 SHELL ["bash", "-Eeuo", "pipefail", "-xc"]
 

Dockerfile.test

@@ -1,4 +1,4 @@
-FROM golang:1.20-bullseye
+FROM golang:1.21-bookworm
 
 SHELL ["bash", "-Eeuo", "pipefail", "-xc"]
 

bashbrew.sh

@@ -6,11 +6,11 @@ set -Eeuo pipefail
 dir="$(readlink -f "$BASH_SOURCE")"
 dir="$(dirname "$dir")"
 
-: "${CGO_ENABLED:=0}"
-export GO111MODULE=on CGO_ENABLED
+: "${CGO_ENABLED=0}" "${GOTOOLCHAIN=local}"
+export CGO_ENABLED GOTOOLCHAIN
 (
 	cd "$dir"
-	go build -o bin/bashbrew ./cmd/bashbrew > /dev/null
+	go build -trimpath -o bin/bashbrew ./cmd/bashbrew > /dev/null
 )
 
 exec "$dir/bin/bashbrew" "$@"

cmd/bashbrew/docker.go

@@ -1,7 +1,6 @@
 package main
 
 import (
-	"bufio"
 	"bytes"
 	"crypto/sha256"
 	"encoding/hex"
@@ -10,21 +9,14 @@ import (
 	"os"
 	"os/exec"
 	"path"
-	"strconv"
 	"strings"
+	"sync"
 
 	"github.com/docker-library/bashbrew/manifest"
+	"github.com/docker-library/bashbrew/pkg/dockerfile"
 	"github.com/urfave/cli"
 )
 
-type dockerfileMetadata struct {
-	StageFroms     []string          // every image "FROM" instruction value (or the parent stage's FROM value in the case of a named stage)
-	StageNames     []string          // the name of any named stage (in order)
-	StageNameFroms map[string]string // map of stage names to FROM values (or the parent stage's FROM value in the case of a named stage), useful for resolving stage names to FROM values
-
-	Froms []string // every "FROM" or "COPY --from=xxx" value (minus named and/or numbered stages in the case of "--from=")
-}
-
 // this returns the "FROM" value for the last stage (which essentially determines the "base" for the final published image)
 func (r Repo) ArchLastStageFrom(arch string, entry *manifest.Manifest2822Entry) (string, error) {
 	dockerfileMeta, err := r.archDockerfileMetadata(arch, entry)
@@ -46,27 +38,25 @@ func (r Repo) ArchDockerFroms(arch string, entry *manifest.Manifest2822Entry) ([
 	return dockerfileMeta.Froms, nil
 }
 
-func (r Repo) dockerfileMetadata(entry *manifest.Manifest2822Entry) (*dockerfileMetadata, error) {
+func (r Repo) dockerfileMetadata(entry *manifest.Manifest2822Entry) (dockerfile.Metadata, error) {
 	return r.archDockerfileMetadata(arch, entry)
 }
 
-var dockerfileMetadataCache = map[string]*dockerfileMetadata{}
+var (
+	dockerfileMetadataCache   = map[string]dockerfile.Metadata{}
+	scratchDockerfileMetadata = sync.OnceValues(func() (dockerfile.Metadata, error) {
+		return dockerfile.Parse(`FROM scratch`)
+	})
+)
 
-func (r Repo) archDockerfileMetadata(arch string, entry *manifest.Manifest2822Entry) (*dockerfileMetadata, error) {
+func (r Repo) archDockerfileMetadata(arch string, entry *manifest.Manifest2822Entry) (dockerfile.Metadata, error) {
 	if builder := entry.ArchBuilder(arch); builder == "oci-import" {
-		return &dockerfileMetadata{
-			StageFroms: []string{
-				"scratch",
-			},
-			Froms: []string{
-				"scratch",
-			},
-		}, nil
+		return scratchDockerfileMetadata()
 	}
 
 	commit, err := r.fetchGitRepo(arch, entry)
 	if err != nil {
-		return nil, cli.NewMultiError(fmt.Errorf("failed fetching Git repo for arch %q from entry %q", arch, entry.String()), err)
+		return dockerfile.Metadata{}, cli.NewMultiError(fmt.Errorf("failed fetching Git repo for arch %q from entry %q", arch, entry.String()), err)
 	}
 
 	dockerfileFile := path.Join(entry.ArchDirectory(arch), entry.ArchFile(arch))
@@ -79,116 +69,20 @@ func (r Repo) archDockerfileMetadata(arch string, entry *manifest.Manifest2822En
 		return meta, nil
 	}
 
-	dockerfile, err := gitShow(commit, dockerfileFile)
+	df, err := gitShow(commit, dockerfileFile)
 	if err != nil {
-		return nil, cli.NewMultiError(fmt.Errorf(`failed "git show" for %q from commit %q`, dockerfileFile, commit), err)
+		return dockerfile.Metadata{}, cli.NewMultiError(fmt.Errorf(`failed "git show" for %q from commit %q`, dockerfileFile, commit), err)
 	}
 
-	meta, err := parseDockerfileMetadata(dockerfile)
+	meta, err := dockerfile.Parse(df)
 	if err != nil {
-		return nil, cli.NewMultiError(fmt.Errorf(`failed parsing Dockerfile metadata for %q from commit %q`, dockerfileFile, commit), err)
+		return dockerfile.Metadata{}, cli.NewMultiError(fmt.Errorf(`failed parsing Dockerfile metadata for %q from commit %q`, dockerfileFile, commit), err)
 	}
 
 	dockerfileMetadataCache[cacheKey] = meta
 	return meta, nil
 }
 
-func parseDockerfileMetadata(dockerfile string) (*dockerfileMetadata, error) {
-	meta := &dockerfileMetadata{
-		// panic: assignment to entry in nil map
-		StageNameFroms: map[string]string{},
-		// (nil slices work fine)
-	}
-
-	scanner := bufio.NewScanner(strings.NewReader(dockerfile))
-	for scanner.Scan() {
-		line := strings.TrimSpace(scanner.Text())
-
-		if line == "" {
-			// ignore blank lines
-			continue
-		}
-
-		if line[0] == '#' {
-			// TODO handle "escape" parser directive
-			// TODO handle "syntax" parser directive -- explode appropriately (since custom syntax invalidates our Dockerfile parsing)
-			// ignore comments
-			continue
-		}
-
-		// handle line continuations
-		// (TODO see note above regarding "escape" parser directive)
-		for line[len(line)-1] == '\\' && scanner.Scan() {
-			nextLine := strings.TrimSpace(scanner.Text())
-			if nextLine == "" || nextLine[0] == '#' {
-				// ignore blank lines and comments
-				continue
-			}
-			line = line[0:len(line)-1] + nextLine
-		}
-
-		fields := strings.Fields(line)
-		if len(fields) < 1 {
-			// must be a much more complex empty line??
-			continue
-		}
-		instruction := strings.ToUpper(fields[0])
-
-		// TODO balk at ARG / $ in from values
-
-		switch instruction {
-		case "FROM":
-			from := fields[1]
-
-			if stageFrom, ok := meta.StageNameFroms[from]; ok {
-				// if this is a valid stage name, we should resolve it back to the original FROM value of that previous stage (we don't care about inter-stage dependencies for the purposes of either tag dependency calculation or tag building -- just how many there are and what external things they require)
-				from = stageFrom
-			}
-
-			// make sure to add ":latest" if it's implied
-			from = latestizeRepoTag(from)
-
-			meta.StageFroms = append(meta.StageFroms, from)
-			meta.Froms = append(meta.Froms, from)
-
-			if len(fields) == 4 && strings.ToUpper(fields[2]) == "AS" {
-				stageName := fields[3]
-				meta.StageNames = append(meta.StageNames, stageName)
-				meta.StageNameFroms[stageName] = from
-			}
-		case "COPY":
-			for _, arg := range fields[1:] {
-				if !strings.HasPrefix(arg, "--") {
-					// doesn't appear to be a "flag"; time to bail!
-					break
-				}
-				if !strings.HasPrefix(arg, "--from=") {
-					// ignore any flags we're not interested in
-					continue
-				}
-				from := arg[len("--from="):]
-
-				if stageFrom, ok := meta.StageNameFroms[from]; ok {
-					// see note above regarding stage names in FROM
-					from = stageFrom
-				} else if stageNumber, err := strconv.Atoi(from); err == nil && stageNumber < len(meta.StageFroms) {
-					// must be a stage number, we should resolve it too
-					from = meta.StageFroms[stageNumber]
-				}
-
-				// make sure to add ":latest" if it's implied
-				from = latestizeRepoTag(from)
-
-				meta.Froms = append(meta.Froms, from)
-			}
-		}
-	}
-	if err := scanner.Err(); err != nil {
-		return nil, err
-	}
-	return meta, nil
-}
-
 func (r Repo) DockerCacheName(entry *manifest.Manifest2822Entry) (string, error) {
 	cacheHash, err := r.dockerCacheHash(entry)
 	if err != nil {

cmd/bashbrew/repo.go

@@ -6,7 +6,6 @@ import (
 	"path"
 	"path/filepath"
 	"sort"
-	"strings"
 
 	"github.com/docker-library/bashbrew/manifest"
 )
@@ -39,13 +38,6 @@ func repos(all bool, args ...string) ([]string, error) {
 	return ret, nil
 }
 
-func latestizeRepoTag(repoTag string) string {
-	if repoTag != "scratch" && strings.IndexRune(repoTag, ':') < 0 {
-		return repoTag + ":latest"
-	}
-	return repoTag
-}
-
 type Repo struct {
 	RepoName string
 	TagName  string

go.mod

@@ -1,6 +1,6 @@
 module github.com/docker-library/bashbrew
 
-go 1.20
+go 1.21
 
 require (
 	github.com/containerd/containerd v1.6.19

go.sum

@@ -617,6 +617,7 @@ github.com/prometheus/procfs v0.1.3/go.mod h1:lV6e/gmhEcM9IjHGsFOCxxuZ+z1YqCvr4O
 github.com/prometheus/procfs v0.2.0/go.mod h1:lV6e/gmhEcM9IjHGsFOCxxuZ+z1YqCvr4OA4YeYWdaU=
 github.com/prometheus/procfs v0.6.0/go.mod h1:cz+aTbrPOrUb4q7XlbU9ygM+/jj0fzG6c1xBZuNvfVA=
 github.com/prometheus/procfs v0.7.3 h1:4jVXhlkAyzOScmCkXBTOLRLTz8EeU+eyjrwB/EPq0VU=
+github.com/prometheus/procfs v0.7.3/go.mod h1:cz+aTbrPOrUb4q7XlbU9ygM+/jj0fzG6c1xBZuNvfVA=
 github.com/prometheus/tsdb v0.7.1/go.mod h1:qhTCs0VvXwvX/y3TZrWD7rabWM+ijKTux40TwIPHuXU=
 github.com/rogpeppe/fastuuid v0.0.0-20150106093220-6724a57986af/go.mod h1:XWv6SoW27p1b0cqNHllgS5HIMJraePCO15w5zCzIWYg=
 github.com/rogpeppe/fastuuid v1.2.0/go.mod h1:jVj6XXZzXRy/MSR5jhDC/2q6DgLz+nrA6LYCDYWNEvQ=