🐛 Fix regex for matching RPM packages with unorthodox vendor name (#4726

)

* Modify regex pattern for SuSE vendor name.

Signed-off-by: Vasil Sirakov <[email protected]>

* Trim out the HTML tag altogether and update the unit tests.

Signed-off-by: Vasil Sirakov <[email protected]>

* Fix unit test name.

Signed-off-by: Vasil Sirakov <[email protected]>

* Cleanup vendor name also in static query code path.

Signed-off-by: Vasil Sirakov <[email protected]>

---------

Signed-off-by: Vasil Sirakov <[email protected]>

providers/os/resources/packages/rpm_packages.go

@@ -30,7 +30,7 @@ const (
 	RpmPkgFormat = "rpm"
 )
 
-var RPM_REGEX = regexp.MustCompile(`^([\w-+]*)\s(\d*|\(none\)):([\w\d-+.:]+)\s([\w\d]*|\(none\))__([\w\d\s,\.]+)__(.*)$`)
+var RPM_REGEX = regexp.MustCompile(`^([\w-+]*)\s(\d*|\(none\)):([\w\d-+.:]+)\s([\w\d]*|\(none\))__([\w\d\s,/<>:\.]+)__(.*)$`)
 
 // ParseRpmPackages parses output from:
 // rpm -qa --queryformat '%{NAME} %{EPOCHNUM}:%{VERSION}-%{RELEASE} %{ARCH}__%{VENDOR}__%{SUMMARY}\n'
@@ -58,7 +58,10 @@ func ParseRpmPackages(pf *inventory.Platform, input io.Reader) []Package {
 			if arch == "(none)" {
 				arch = ""
 			}
-			pkg := newRpmPackage(pf, name, version, arch, epoch, m[5], m[6])
+
+			vendor := cleanupVendorName(m[5])
+
+			pkg := newRpmPackage(pf, name, version, arch, epoch, vendor, m[6])
 			pkg.FilesAvailable = PkgFilesAsync // when we use commands we need to fetch the files async
 			pkgs = append(pkgs, pkg)
 
@@ -97,8 +100,19 @@ func newRpmPackage(pf *inventory.Platform, name, version, arch, epoch, vendor, d
 	}
 }
 
+// matches a closed pair of angle brackets with any number of characters inside.
+var CLEANUP_VENDOR_REGEX = regexp.MustCompile(`<.*>`)
+
+// it is possible for the vendor name to contain a HTML tag with a website inside, e.g. SUSE.
+// we remove it because it is not necessary and later causes troubles for the CPE generation.
+// this assumes angle brackets are not used anywhere else in the names of vendors which is already the case.
+func cleanupVendorName(vendor string) string {
+	cleaned := CLEANUP_VENDOR_REGEX.ReplaceAllString(vendor, "")
+	return strings.TrimRight(cleaned, " ")
+}
+
 // RpmPkgManager is the package manager for Redhat, CentOS, Oracle, Photon and Suse
-// it support two modes: runtime where the rpm command is available and static analysis for images (e.g. container tar)
+// it supports two modes: runtime where the rpm command is available and static analysis for images (e.g. container tar)
 // If the RpmPkgManager is used in static mode, it extracts the rpm database from the system and copies it to the local
 // filesystem to run a local rpm command to extract the data. The static analysis is always slower than using the running
 // one since more data need to copied. Therefore the runtime check should be preferred over the static analysis
@@ -278,7 +292,7 @@ func (rpm *RpmPkgManager) staticList() ([]Package, error) {
 			version = version + "-" + pkg.Release
 		}
 
-		rpmPkg := newRpmPackage(rpm.platform, pkg.Name, version, pkg.Arch, epoch, pkg.Vendor, pkg.Summary)
+		rpmPkg := newRpmPackage(rpm.platform, pkg.Name, version, pkg.Arch, epoch, cleanupVendorName(pkg.Vendor), pkg.Summary)
 
 		// determine all files attached
 		records := []FileRecord{}

providers/os/resources/packages/rpm_packages_test.go

@@ -287,3 +287,88 @@ func TestPhoton4ImageParser(t *testing.T) {
 	}
 	assert.Equal(t, p, findPkg(m, p.Name), p.Name)
 }
+
+// ensure that the tag in the SuSE vendor name is parsed correctly
+func TestSuSEParser(t *testing.T) {
+	epoch := int(0)
+	pkgList := []*rpmdb.PackageInfo{
+		{
+			Name:    "grep",
+			Epoch:   &epoch,
+			Version: "3.1",
+			Release: "150000.4.6.1",
+			Arch:    "x86_64",
+			Vendor:  "SUSE LLC <https://www.suse.com/>",
+			Summary: "Print lines matching a pattern",
+		},
+	}
+
+	var packageList bytes.Buffer
+	for _, pkg := range pkgList {
+		packageList.WriteString(fmt.Sprintf("%s %d:%s-%s %s__%s__%s\n", pkg.Name, pkg.EpochNum(), pkg.Version, pkg.Release, pkg.Arch, pkg.Vendor, pkg.Summary))
+	}
+
+	pf := &inventory.Platform{
+		Name:    "suse",
+		Version: "15.6",
+		Arch:    "x86_64",
+		Family:  []string{"linux", "unix", "os"},
+		Labels: map[string]string{
+			"distro-id": "suse",
+		},
+	}
+
+	m := ParseRpmPackages(pf, &packageList)
+	assert.Equal(t, 1, len(m), "detected the right amount of packages")
+
+	p := Package{
+		Name:    "grep",
+		Version: "3.1-150000.4.6.1",
+		// Note that the tag <https://suse.com/> has been removed.
+		Vendor:      "SUSE LLC",
+		Arch:        "x86_64",
+		Description: "Print lines matching a pattern",
+		PUrl:        "pkg:rpm/suse/[email protected]?arch=x86_64&distro=suse-15.6",
+		CPEs: []string{
+			"cpe:2.3:a:suse_llc:grep:3.1-150000.4.6.1:*:*:*:*:*:x86_64:*",
+			"cpe:2.3:a:suse_llc:grep:3.1-150000.4:*:*:*:*:*:x86_64:*",
+			"cpe:2.3:a:suse_llc:grep:3.1:*:*:*:*:*:x86_64:*",
+			"cpe:2.3:a:suse_llc:grep:3.1-150000.4.6.1:*:*:*:*:*:*:*",
+			"cpe:2.3:a:suse_llc:grep:3.1-150000.4:*:*:*:*:*:*:*",
+			"cpe:2.3:a:suse_llc:grep:3.1:*:*:*:*:*:*:*",
+		},
+		Format:         RpmPkgFormat,
+		FilesAvailable: PkgFilesAsync,
+	}
+	assert.Equal(t, p, m[0], p.Name)
+}
+
+func TestVendorNameCleanup(t *testing.T) {
+	vendorFromRpm := "SUSE LLC"
+	actual := cleanupVendorName(vendorFromRpm)
+	require.Equal(t, "SUSE LLC", actual)
+
+	vendorFromRpm = "SUSE LLC<https://suse.com/>"
+	actual = cleanupVendorName(vendorFromRpm)
+	require.Equal(t, "SUSE LLC", actual)
+
+	vendorFromRpm = "SUSE LLC <https://suse.com/>"
+	actual = cleanupVendorName(vendorFromRpm)
+	require.Equal(t, "SUSE LLC", actual)
+
+	vendorFromRpm = "SUSE LLC     <https://suse.com/>"
+	actual = cleanupVendorName(vendorFromRpm)
+	require.Equal(t, "SUSE LLC", actual)
+
+	vendorFromRpm = "SUSE LLC <abc><def>"
+	actual = cleanupVendorName(vendorFromRpm)
+	require.Equal(t, "SUSE LLC", actual)
+
+	vendorFromRpm = "SUSE LLC <>"
+	actual = cleanupVendorName(vendorFromRpm)
+	require.Equal(t, "SUSE LLC", actual)
+
+	vendorFromRpm = "SUSE LLC <<>>"
+	actual = cleanupVendorName(vendorFromRpm)
+	require.Equal(t, "SUSE LLC", actual)
+}