Unpin fixture dependencies that will always float (#4495)

* unpin dependencies for package dedup case

Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>

* prevent make from blocking tests

Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>

* add check

Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>

---------

Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>

.github/actions/bootstrap/action.yaml

@@ -45,7 +45,10 @@ runs:
     - name: Install project tools
       shell: bash
       if: inputs.tools == 'true'
-      run: make tools
+      run: |
+        make tools
+        .tool/binny list
+        .tool/binny check
 
     - name: Install go dependencies
       if: inputs.go-dependencies == 'true'

.github/workflows/test-fixture-cache-publish.yaml

@@ -30,7 +30,9 @@ jobs:
           download-test-fixture-cache: false
 
       - name: Run all tests
-        run: make test
+        run: |
+          make list-tasks-targets
+          make --debug=verbose test
         env:
           # we want to rebuild the cache with no previous state
           DOWNLOAD_TEST_FIXTURE_CACHE: "false"

Makefile

@@ -36,7 +36,7 @@ make-default: $(TASK)
 	@$(TASK)
 
 # for those of us that can't seem to kick the habit of typing `make ...` lets wrap the superior `task` tool
-TASKS := $(shell bash -c "test -f $(TASK) && $(TASK) -l | grep '^\* ' | cut -d' ' -f2 | tr -d ':' | tr '\n' ' '" ) $(shell bash -c "test -f $(TASK) && $(TASK) -l | grep 'aliases:' | cut -d ':' -f 3 | tr '\n' ' ' | tr -d ','")
+TASKS := $(shell bash -c "test -f $(TASK) && NO_COLOR=1 $(TASK) -l | grep '^\* ' | cut -d' ' -f2 | tr -d ':' | tr '\n' ' '" ) $(shell bash -c "test -f $(TASK) && NO_COLOR=1 $(TASK) -l | grep 'aliases:' | cut -d ':' -f 3 | tr '\n' ' ' | tr -d ','")
 
 .PHONY: $(TASKS)
 $(TASKS): $(TASK)

cmd/syft/internal/test/integration/package_deduplication_test.go

@@ -1,7 +1,6 @@
 package integration
 
 import (
-	"fmt"
 	"testing"
 
 	"github.com/stretchr/testify/assert"
@@ -11,47 +10,52 @@ import (
 )
 
 func TestPackageDeduplication(t *testing.T) {
+	// this test verifies that package deduplication works correctly across layers.
+	// The test fixture installs/upgrades packages in multiple stages, creating
+	// duplicate RPM DB entries across layers. Without deduplication, we'd see ~600 packages.
+	//
+	// Note: we index by package name (not name-version) to be resilient to Rocky Linux
+	// repo updates. Location counts are summed across all versions of each package.
 	tests := []struct {
 		scope         source.Scope
 		packageCount  int
-		instanceCount map[string]int
+		instanceCount map[string]int // how many distinct package instances (by name)
-		locationCount map[string]int
+		locationCount map[string]int // total locations across ALL versions of each package
 	}{
 		{
 			scope:        source.AllLayersScope,
-			packageCount: 175, // without deduplication this would be ~600
+			packageCount: 176, // without deduplication this would be ~600
 			instanceCount: map[string]int{
 				"basesystem":   1,
 				"wget":         1,
-				"curl-minimal": 2, // upgraded in the image
+				"curl-minimal": 2, // base + upgraded (2 different versions)
 				"vsftpd":       1,
-				"httpd":        1, // rpm, - we exclude binary
+				"httpd":        1,
 			},
 			locationCount: map[string]int{
-				"basesystem-11-13.el9":               5, // in all layers
+				"basesystem":   5, // in all layers
-				"curl-minimal-7.76.1-26.el9_3.2.0.1": 2, // base + wget layer
+				"curl-minimal": 5, // total across both versions (2 + 3)
-				"curl-minimal-7.76.1-31.el9_6.1":     3, // curl upgrade layer + all above layers
+				"wget":         4, // wget + all above layers
-				"wget-1.21.1-8.el9_4":                4, // wget + all above layers
+				"vsftpd":       2, // vsftpd + all above layers
-				"vsftpd-3.0.5-6.el9":                 2, // vsftpd + all above layers
+				"httpd":        1, // last layer
-				"httpd-2.4.62-4.el9_6.4":             1, // last layer
 			},
 		},
 		{
 			scope:        source.SquashedScope,
-			packageCount: 169,
+			packageCount: 170,
 			instanceCount: map[string]int{
 				"basesystem":   1,
 				"wget":         1,
-				"curl-minimal": 1, // upgraded, but the most recent
+				"curl-minimal": 1, // deduped to latest
 				"vsftpd":       1,
-				"httpd":        1, // rpm, binary is now excluded by overlap
+				"httpd":        1,
 			},
 			locationCount: map[string]int{
-				"basesystem-11-13.el9":           1,
+				"basesystem":   1,
-				"curl-minimal-7.76.1-31.el9_6.1": 1, // upgrade
+				"curl-minimal": 1,
-				"wget-1.21.1-8.el9_4":            1,
+				"wget":         1,
-				"vsftpd-3.0.5-6.el9":             1,
+				"vsftpd":       1,
-				"httpd-2.4.62-4.el9_6.4":         1,
+				"httpd":        1,
 			},
 		},
 	}
@@ -59,35 +63,32 @@ func TestPackageDeduplication(t *testing.T) {
 	for _, tt := range tests {
 		t.Run(string(tt.scope), func(t *testing.T) {
 			sbom, _ := catalogFixtureImage(t, "image-vertical-package-dups", tt.scope)
+
+			// verify binary packages have names
 			for _, p := range sbom.Artifacts.Packages.Sorted() {
 				if p.Type == pkg.BinaryPkg {
 					assert.NotEmpty(t, p.Name)
 				}
 			}
 
+			// verify exact package count
 			assert.Equal(t, tt.packageCount, sbom.Artifacts.Packages.PackageCount())
-			for name, expectedInstanceCount := range tt.instanceCount {
+
+			// verify instance count by package name
+			for name, expectedCount := range tt.instanceCount {
 				pkgs := sbom.Artifacts.Packages.PackagesByName(name)
+				assert.Len(t, pkgs, expectedCount, "unexpected instance count for %s", name)
+			}
 
-				// with multiple packages with the same name, something is wrong (or this is the wrong fixture)
+			// verify total location count across all versions of each package
-				if assert.Len(t, pkgs, expectedInstanceCount, "unexpected package count for %s", name) {
+			for name, expectedLocCount := range tt.locationCount {
+				pkgs := sbom.Artifacts.Packages.PackagesByName(name)
+				totalLocations := 0
 				for _, p := range pkgs {
-						nameVersion := fmt.Sprintf("%s-%s", name, p.Version)
+					totalLocations += len(p.Locations.ToSlice())
-						expectedLocationCount, ok := tt.locationCount[nameVersion]
-						if !ok {
-							t.Errorf("missing name-version: %s", nameVersion)
-							continue
 				}
-
+				assert.Equal(t, expectedLocCount, totalLocations, "unexpected total location count for %s", name)
-						// we should see merged locations (assumption, there was 1 location for each package)
-						assert.Len(t, p.Locations.ToSlice(), expectedLocationCount, "unexpected location count for %s", nameVersion)
-
-						// all paths should match
-						assert.Len(t, p.Locations.CoordinateSet().Paths(), 1, "unexpected location count for %s", nameVersion)
 			}
-				}
-			}
-
 		})
 	}
 }

cmd/syft/internal/test/integration/test-fixtures/image-vertical-package-dups/Dockerfile

@@ -7,16 +7,16 @@ FROM --platform=linux/amd64 rockylinux:9.3.20231119@sha256:d644d203142cd5b54ad2a
 # copying the RPM DB from each stage to a final stage in separate layers. This will result in a much smaller image.
 
 FROM base AS stage1
-RUN dnf install -y wget-1.21.1-8.el9_4
+RUN dnf install -y wget
 
 FROM stage1 AS stage2
-RUN dnf update -y curl-minimal-7.76.1-31.el9_6.1
+RUN dnf update -y curl-minimal
 
 FROM stage2 AS stage3
-RUN dnf install -y vsftpd-3.0.5-6.el9
+RUN dnf install -y vsftpd
 
 FROM stage3 AS stage4
-RUN dnf install -y httpd-2.4.62-4.el9_6.4
+RUN dnf install -y httpd
 
 FROM scratch