add squashed with all layers resolver

Signed-off-by: Alex Goodman <alex.goodman@anchore.com>

syft/pkg/package.go

@@ -59,6 +59,8 @@ func (p *Package) merge(other Package) error {
 		return fmt.Errorf("cannot merge packages with different IDs: %q vs %q", p.id, other.id)
 	}
 
+	log.WithFields("id", p.id, "purl", p.PURL).Trace("merging similar packages")
+
 	if p.PURL != other.PURL {
 		log.Warnf("merging packages have with different pURLs: %q=%q vs %q=%q", p.id, p.PURL, other.id, other.PURL)
 	}

syft/source/image_all_layers_resolver.go

@@ -18,8 +18,8 @@ type imageAllLayersResolver struct {
 	layers []int
 }
 
-// newAllLayersResolver returns a new resolver from the perspective of all image layers for the given image.
+// newImageAllLayersResolver returns a new resolver from the perspective of all image layers for the given image.
-func newAllLayersResolver(img *image.Image) (*imageAllLayersResolver, error) {
+func newImageAllLayersResolver(img *image.Image) (*imageAllLayersResolver, error) {
 	if len(img.Layers) == 0 {
 		return nil, fmt.Errorf("the image does not contain any layers")
 	}
@@ -202,7 +202,7 @@ func (r *imageAllLayersResolver) FileContentsByLocation(location Location) (io.R
 		return nil, fmt.Errorf("cannot read contents of non-file %q", location.ref.RealPath)
 	}
 
-	return r.img.FileContentsByRef(location.ref)
+	return r.img.OpenReference(location.ref)
 }
 
 func (r *imageAllLayersResolver) FilesByMIMEType(types ...string) ([]Location, error) {

syft/source/image_all_layers_resolver_test.go

@@ -91,7 +91,7 @@ func TestAllLayersResolver_FilesByPath(t *testing.T) {
 		t.Run(c.name, func(t *testing.T) {
 			img := imagetest.GetFixtureImage(t, "docker-archive", "image-symlinks")
 
-			resolver, err := newAllLayersResolver(img)
+			resolver, err := newImageAllLayersResolver(img)
 			if err != nil {
 				t.Fatalf("could not create resolver: %+v", err)
 			}
@@ -205,7 +205,7 @@ func TestAllLayersResolver_FilesByGlob(t *testing.T) {
 		t.Run(c.name, func(t *testing.T) {
 			img := imagetest.GetFixtureImage(t, "docker-archive", "image-symlinks")
 
-			resolver, err := newAllLayersResolver(img)
+			resolver, err := newImageAllLayersResolver(img)
 			if err != nil {
 				t.Fatalf("could not create resolver: %+v", err)
 			}
@@ -257,7 +257,7 @@ func Test_imageAllLayersResolver_FilesByMIMEType(t *testing.T) {
 		t.Run(test.fixtureName, func(t *testing.T) {
 			img := imagetest.GetFixtureImage(t, "docker-archive", test.fixtureName)
 
-			resolver, err := newAllLayersResolver(img)
+			resolver, err := newImageAllLayersResolver(img)
 			assert.NoError(t, err)
 
 			locations, err := resolver.FilesByMIMEType(test.mimeType)
@@ -274,7 +274,7 @@ func Test_imageAllLayersResolver_FilesByMIMEType(t *testing.T) {
 func Test_imageAllLayersResolver_hasFilesystemIDInLocation(t *testing.T) {
 	img := imagetest.GetFixtureImage(t, "docker-archive", "image-duplicate-path")
 
-	resolver, err := newAllLayersResolver(img)
+	resolver, err := newImageAllLayersResolver(img)
 	assert.NoError(t, err)
 
 	locations, err := resolver.FilesByMIMEType("text/plain")
@@ -334,7 +334,7 @@ func TestAllLayersImageResolver_FilesContents(t *testing.T) {
 		t.Run(test.name, func(t *testing.T) {
 			img := imagetest.GetFixtureImage(t, "docker-archive", "image-symlinks")
 
-			resolver, err := newAllLayersResolver(img)
+			resolver, err := newImageAllLayersResolver(img)
 			assert.NoError(t, err)
 
 			refs, err := resolver.FilesByPath(test.fixture)
@@ -361,7 +361,7 @@ func TestAllLayersImageResolver_FilesContents_errorOnDirRequest(t *testing.T) {
 
 	img := imagetest.GetFixtureImage(t, "docker-archive", "image-symlinks")
 
-	resolver, err := newAllLayersResolver(img)
+	resolver, err := newImageAllLayersResolver(img)
 	assert.NoError(t, err)
 
 	var dirLoc *Location
@@ -675,7 +675,7 @@ func Test_imageAllLayersResolver_resolvesLinks(t *testing.T) {
 
 			img := imagetest.GetFixtureImage(t, "docker-archive", "image-symlinks")
 
-			resolver, err := newAllLayersResolver(img)
+			resolver, err := newImageAllLayersResolver(img)
 			assert.NoError(t, err)
 
 			actual := test.runner(resolver)
@@ -689,7 +689,7 @@ func Test_imageAllLayersResolver_resolvesLinks(t *testing.T) {
 func TestAllLayersResolver_AllLocations(t *testing.T) {
 	img := imagetest.GetFixtureImage(t, "docker-archive", "image-files-deleted")
 
-	resolver, err := newAllLayersResolver(img)
+	resolver, err := newImageAllLayersResolver(img)
 	assert.NoError(t, err)
 
 	paths := strset.New()

syft/source/image_squash_resolver.go

@@ -168,7 +168,7 @@ func (r *imageSquashResolver) FileContentsByLocation(location Location) (io.Read
 		return nil, fmt.Errorf("unable to get file contents for directory: %+v", location)
 	}
 
-	return r.img.FileContentsByRef(location.ref)
+	return r.img.OpenReference(location.ref)
 }
 
 func (r *imageSquashResolver) AllLocations() <-chan Location {

syft/source/image_squash_with_all_layers_resolver.go

@@ -0,0 +1,98 @@
+package source
+
+import (
+	"github.com/anchore/stereoscope/pkg/image"
+	"io"
+)
+
+var _ FileResolver = (*imageSquashWithAllLayersResolver)(nil)
+
+// imageSquashWithAllLayersResolver acts like a squash resolver, but additionally returns all paths in earlier layers
+// that have been added/modified (like the all-layers resolver).
+type imageSquashWithAllLayersResolver struct {
+	squashed  *imageSquashResolver
+	allLayers *imageAllLayersResolver
+}
+
+// newImageSquashWithAllLayersResolver returns a new resolver from the perspective of the squashed representation for
+// the given image, but additionally returns all instances of a path that have been added/modified.
+func newImageSquashWithAllLayersResolver(img *image.Image) (*imageSquashWithAllLayersResolver, error) {
+	squashed, err := newImageSquashResolver(img)
+	if err != nil {
+		return nil, err
+	}
+
+	allLayers, err := newImageAllLayersResolver(img)
+	if err != nil {
+		return nil, err
+	}
+
+	return &imageSquashWithAllLayersResolver{
+		squashed:  squashed,
+		allLayers: allLayers,
+	}, nil
+}
+
+func (i imageSquashWithAllLayersResolver) FileContentsByLocation(location Location) (io.ReadCloser, error) {
+	return i.squashed.FileContentsByLocation(location)
+}
+
+func (i imageSquashWithAllLayersResolver) HasPath(s string) bool {
+	return i.squashed.HasPath(s)
+}
+
+func (i imageSquashWithAllLayersResolver) filterLocations(locations []Location, err error) ([]Location, error) {
+	if err != nil {
+		return locations, err
+	}
+	var ret []Location
+	for _, l := range locations {
+		if i.squashed.HasPath(l.RealPath) {
+			// not only should the real path to the file exist, but the way we took to get there should also exist
+			// (e.g. if we are looking for /etc/passwd, but the real path is /etc/passwd -> /etc/passwd-1, then we should
+			// make certain that /etc/passwd-1 exists)
+			if l.VirtualPath != "" && !i.squashed.HasPath(l.VirtualPath) {
+				continue
+			}
+			ret = append(ret, l)
+		}
+	}
+	return ret, nil
+}
+
+func (i imageSquashWithAllLayersResolver) FilesByPath(paths ...string) ([]Location, error) {
+	return i.filterLocations(i.allLayers.FilesByPath(paths...))
+}
+
+func (i imageSquashWithAllLayersResolver) FilesByGlob(patterns ...string) ([]Location, error) {
+	return i.filterLocations(i.allLayers.FilesByGlob(patterns...))
+}
+
+func (i imageSquashWithAllLayersResolver) FilesByMIMEType(types ...string) ([]Location, error) {
+	return i.filterLocations(i.allLayers.FilesByMIMEType(types...))
+}
+
+func (i imageSquashWithAllLayersResolver) RelativeFileByPath(l Location, path string) *Location {
+	if !i.squashed.HasPath(path) {
+		return nil
+	}
+	return i.allLayers.RelativeFileByPath(l, path)
+}
+
+func (i imageSquashWithAllLayersResolver) AllLocations() <-chan Location {
+	var ret = make(chan Location)
+	go func() {
+		defer close(ret)
+		for l := range i.allLayers.AllLocations() {
+			if i.squashed.HasPath(l.RealPath) {
+				ret <- l
+			}
+		}
+	}()
+
+	return ret
+}
+
+func (i imageSquashWithAllLayersResolver) FileMetadataByLocation(location Location) (FileMetadata, error) {
+	return fileMetadataByLocation(i.squashed.img, location)
+}

syft/source/scope.go

@@ -10,14 +10,17 @@ const (
 	UnknownScope Scope = "UnknownScope"
 	// SquashedScope indicates to only catalog content visible from the squashed filesystem representation (what can be seen only within the container at runtime)
 	SquashedScope Scope = "Squashed"
-	// AllLayersScope indicates to catalog content on all layers, irregardless if it is visible from the container at runtime.
+	// AllLayersScope indicates to catalog content on all layers, regardless if it is visible from the container at runtime.
 	AllLayersScope Scope = "AllLayers"
+	// SquashedWithAllLayersScope indicates to catalog content on all layers, but only include content visible from the squashed filesystem representation.
+	SquashedWithAllLayersScope Scope = "SquashedWithAllLayers"
 )
 
 // AllScopes is a slice containing all possible scope options
 var AllScopes = []Scope{
 	SquashedScope,
 	AllLayersScope,
+	SquashedWithAllLayersScope,
 }
 
 // ParseScope returns a scope as indicated from the given string.
@@ -27,6 +30,8 @@ func ParseScope(userStr string) Scope {
 		return SquashedScope
 	case "all-layers", strings.ToLower(AllLayersScope.String()):
 		return AllLayersScope
+	case "squashed-with-all-layers", strings.ToLower(SquashedWithAllLayersScope.String()):
+		return SquashedWithAllLayersScope
 	}
 	return UnknownScope
 }

syft/source/source.go

@@ -466,7 +466,9 @@ func (s *Source) FileResolver(scope Scope) (FileResolver, error) {
 		case SquashedScope:
 			resolver, err = newImageSquashResolver(s.Image)
 		case AllLayersScope:
-			resolver, err = newAllLayersResolver(s.Image)
+			resolver, err = newImageAllLayersResolver(s.Image)
+		case SquashedWithAllLayersScope:
+			resolver, err = newImageSquashWithAllLayersResolver(s.Image)
 		default:
 			return nil, fmt.Errorf("bad image scope provided: %+v", scope)
 		}