Add unknown license reference to package dereferencing #2965 #1379

There are unknown license statements like "This file is the same
license as the package django" which refers to a package which this
file is a part of. This is fixed by extending the dereferencing logic
to look for packages that the file belongs to and using the detected
licenses from the package.

Signed-off-by: Ayan Sinha Mahapatra <ayansmahapatra@gmail.com>

src/licensedcode/data/rules/apache-2.0_306.RULE

@@ -1 +1 @@
-This file is distributed under the same license as the Puppet automation framework package.
+This file is distributed under the same license as the {{Puppet}} {{automation}} framework package.

src/licensedcode/data/rules/apache-2.0_306.yml

@@ -1,3 +1,5 @@
 license_expression: apache-2.0
 is_license_notice: yes
 relevance: 100
+referenced_filenames:
+   - package

src/licensedcode/data/rules/bsd-new_1043.RULE

@@ -1,2 +1,2 @@
 This file is distributed under the same license as the
- debian files of the p11-kit package.
+ {{debian}} files of the {{p11-kit}} package.

src/licensedcode/data/rules/bsd-new_1043.yml

@@ -2,3 +2,5 @@ license_expression: bsd-new
 is_license_notice: yes
 relevance: 100
 minimum_coverage: 100
+referenced_filenames:
+   -package

src/licensedcode/data/rules/bsd-new_1044.RULE

@@ -1,2 +1,2 @@
-License: same-as-rest-of-p11kit
- This file is distributed under the same license as the p11-kit package.
+License: {{same-as-rest-of-p11kit}}
+ This file is distributed under the same license as the {{p11-kit}} package.

src/licensedcode/data/rules/bsd-new_1044.yml

@@ -1,3 +1,6 @@
 license_expression: bsd-new
 is_license_notice: yes
 minimum_coverage: 100
+referenced_filenames:
+   - package
+

src/licensedcode/data/rules/free-unknown-package_1.RULE

@@ -0,0 +1 @@
+This file is distributed under the same license as the package.

src/licensedcode/data/rules/free-unknown-package_1.yml

@@ -0,0 +1,7 @@
+license_expression: free-unknown
+is_license_reference: yes
+is_continuous: yes
+relevance: 100
+minimum_coverage: 100
+referenced_filenames:
+    - package

src/licensedcode/data/rules/free-unknown_80.RULE → src/licensedcode/data/rules/free-unknown-package_2.RULE

@@ -1 +1 @@
-This file is distributed under the same license as the PACKAGE package 
+This file is distributed under the same license as the PACKAGE package.

src/licensedcode/data/rules/free-unknown-package_2.yml

@@ -0,0 +1,7 @@
+license_expression: free-unknown
+is_license_reference: yes
+is_continuous: yes
+relevance: 100
+minimum_coverage: 100
+referenced_filenames:
+    - package

src/licensedcode/data/rules/free-unknown-package_3.RULE

@@ -0,0 +1 @@
+This file is distributed under the same license as the DJANGO package.

src/licensedcode/data/rules/free-unknown-package_3.yml

@@ -0,0 +1,7 @@
+license_expression: free-unknown
+is_license_reference: yes
+is_continuous: yes
+relevance: 100
+minimum_coverage: 100
+referenced_filenames:
+    - package

src/licensedcode/data/rules/gpl-2.0-plus_1070.RULE

@@ -0,0 +1 @@
+This file is distributed under the same license as the {{fusiondirectory}}

src/licensedcode/data/rules/gpl-2.0-plus_1070.yml

@@ -0,0 +1,6 @@
+license_expression: gpl-2.0-plus
+is_license_reference: yes
+relevance: 80
+minimum_coverage: 100
+referenced_filenames:
+   - package

src/licensedcode/data/rules/gpl-2.0-plus_31.RULE

@@ -1 +1 @@
-# This file is distributed under the same license as the Pychess package.
+This file is distributed under the same license as the {{Pychess}} package.

src/licensedcode/data/rules/gpl-2.0-plus_31.yml

@@ -2,4 +2,6 @@ license_expression: gpl-2.0-plus
 is_license_reference: yes
 relevance: 80
 minimum_coverage: 100
+referenced_filenames:
+   - package
 notes: In 2006 Pychess used the GPL 2.0. It became GPL 3.0 in 2011

src/licensedcode/data/rules/gpl-2.0-plus_793.RULE

@@ -1,2 +1,2 @@
 GNU GPL:
-This file is distributed under the same license as the dialog package.
+This file is distributed under the same license as the {{dialog}} package.

src/licensedcode/data/rules/gpl-2.0-plus_793.yml

@@ -2,3 +2,5 @@ license_expression: gpl-2.0-plus
 is_license_notice: yes
 relevance: 100
 minimum_coverage: 100
+referenced_filenames:
+   - package

src/licensedcode/data/rules/gpl-2.0_109.RULE

@@ -1 +1 @@
-# This file is distributed under the same license as the cpufrequtils package.
+This file is distributed under the same license as the {{cpufrequtils}} package.

src/licensedcode/data/rules/gpl-2.0_109.yml

@@ -2,4 +2,6 @@ license_expression: gpl-2.0
 is_license_reference: yes
 relevance: 100
 minimum_coverage: 100
+referenced_filenames:
+   - package
 notes: cpufrequtils is gpl-2.0 only

src/licensedcode/data/rules/lgpl-2.1-plus_283.RULE

@@ -1 +1 @@
-This file is distributed under the same license as the dialog package.
+This file is distributed under the same license as the {{dialog}} package.

src/licensedcode/data/rules/lgpl-2.1-plus_283.yml

@@ -2,3 +2,5 @@ license_expression: lgpl-2.1-plus
 is_license_notice: yes
 relevance: 100
 minimum_coverage: 100
+referenced_filenames:
+   - package

src/licensedcode/data/rules/openssl-ssleay_44.RULE

@@ -1 +1 @@
-This file is distributed under the same license as OpenSSL
+This file is distributed under the same license as {{OpenSSL}}

src/licensedcode/data/rules/openssl-ssleay_44.yml

@@ -1,4 +1,6 @@
 license_expression: openssl-ssleay
 is_license_notice: yes
 relevance: 100
+referenced_filenames:
+   - package
 notes: Seen in https://github.com/schmel/openssl-gost/blob/18028d11296f4675276f0814674c5143caaa75e7/engine/gost_grasshopper_mac.c#L3

src/licensedcode/data/rules/ruby_or_lgpl-3.0-plus_3.RULE

@@ -1 +1 @@
-This file is distributed under the same license as gettext.
+This file is distributed under the same license as {{gettext}}

src/licensedcode/data/rules/ruby_or_lgpl-3.0-plus_3.yml

@@ -2,4 +2,6 @@ license_expression: ruby OR lgpl-3.0-plus
 is_license_notice: yes
 relevance: 90
 minimum_coverage: 100
+referenced_filenames:
+   - package
 notes: this is found mostly in Ruby gettext

src/licensedcode/detection.py

@@ -81,6 +81,7 @@ class DetectionCategory(Enum):
     PERFECT_DETECTION = 'perfect-detection'
     UNKNOWN_INTRO_BEFORE_DETECTION = 'unknown-intro-before-detection'
     UNKNOWN_FILE_REFERENCE_LOCAL = 'unknown-file-reference-local'
+    UNKNOWN_REFERENCE_IN_FILE_TO_PACKAGE = 'unknown-reference-in-file-to-package'
     PACKAGE_UNKNOWN_FILE_REFERENCE_LOCAL = 'package-unknown-file-reference-local'
     PACKAGE_ADD_FROM_SIBLING_FILE = 'from-package-sibling-file'
     PACKAGE_ADD_FROM_FILE = 'from-package-file'
@@ -99,6 +100,7 @@ class DetectionRule(Enum):
     FALSE_POSITIVE = 'false-positive'
     UNKNOWN_REFERENCE_TO_LOCAL_FILE = 'unknown-reference-to-local-file' 
     UNKNOWN_INTRO_FOLLOWED_BY_MATCH = 'unknown-intro-followed-by-match'
+    UNKNOWN_REFERENCE_IN_FILE_TO_PACKAGE = 'unknown-reference-in-file-to-package'
     CONTAINED_SAME_LICENSE = 'contained-with-same-license'
     NOTICE_FOLLOWED_BY_TEXT = 'notice-followed-by-text'
     CONTIGUOUS_SAME_LICENSE = 'contiguous-with-same-license'
@@ -443,6 +445,7 @@ def is_correct_detection(license_matches):
     Return True if all the matches in `license_matches` List of LicenseMatch
     are correct license detections.
     """
+    #TODO: Add matches with full match coverage
     matchers = (license_match.matcher for license_match in license_matches)
     return (
         all(matcher in ("1-hash", "1-spdx-id") for matcher in matchers)
@@ -578,7 +581,6 @@ def is_license_clues(license_matches):
     license detection and are mere license clues.
     """
     return not is_correct_detection(license_matches) and (
-        has_unknown_matches(license_matches) or
         is_match_coverage_less_than_threshold(
             license_matches=license_matches,
             threshold=CLUES_MATCH_COVERAGE_THR,
@@ -718,7 +720,11 @@ def get_detected_license_expression(matches, analysis, post_scan=False):
         reasons.append(DetectionRule.UNKNOWN_INTRO_FOLLOWED_BY_MATCH.value)
 
     elif post_scan:
-        if analysis == DetectionCategory.UNKNOWN_FILE_REFERENCE_LOCAL.value:
+        if analysis == DetectionCategory.UNKNOWN_REFERENCE_IN_FILE_TO_PACKAGE.value:
+            matches_for_expression = filter_license_references(matches)
+            reasons.append(DetectionRule.UNKNOWN_REFERENCE_IN_FILE_TO_PACKAGE.value)
+
+        elif analysis == DetectionCategory.UNKNOWN_FILE_REFERENCE_LOCAL.value:
             matches_for_expression = filter_license_references(matches)
             reasons.append(DetectionRule.UNKNOWN_REFERENCE_TO_LOCAL_FILE.value)
 

src/packagedcode/licensing.py

@@ -143,6 +143,87 @@ def add_referenced_license_matches_for_package(resource, codebase, no_licenses):
             yield resource
 
 
+def add_referenced_license_detection_from_package(resource, codebase):
+    """
+    Return an updated ``resource`` saving it in place, after adding new license
+    matches (licenses and license_expressions) following their Rule
+    ``referenced_filenames`` if it is pointing to a package.
+    """
+    if TRACE:
+        logger_debug(f'packagedcode.licensing: add_referenced_license_matches_from_package: resource: {resource.path}')
+
+    if not resource.is_file:
+        return
+
+    license_detections = resource.license_detections
+    if not license_detections:
+        return
+
+    codebase_packages = codebase.attributes.packages
+
+    modified = False
+
+    for detection in license_detections:
+        detection_modified = False
+        matches = detection["matches"]
+        referenced_filenames = get_referenced_filenames(matches)
+        if not referenced_filenames:
+            continue
+
+        has_reference_to_package = any([
+            'package' in referenced_filename
+            for referenced_filename in referenced_filenames
+        ])
+
+        if not has_reference_to_package:
+            continue
+
+        for_packages = resource.for_packages
+        for package_uid in for_packages:
+
+            for codebase_package in codebase_packages:
+                if codebase_package["package_uid"] == package_uid:
+                    break
+
+            pkg_detections = codebase_package["license_detections"]
+            for pkg_detection in pkg_detections:
+                modified = True
+                detection_modified = True
+                matches.extend(pkg_detection["matches"])
+
+        if not detection_modified:
+            continue
+
+        reasons, license_expression = get_detected_license_expression(
+            matches=matches,
+            analysis=DetectionCategory.UNKNOWN_REFERENCE_IN_FILE_TO_PACKAGE.value,
+            post_scan=True,
+        )
+        detection["license_expression"] = str(license_expression)
+        detection["detection_rules"] = reasons
+
+    if modified:
+
+        license_expressions = [
+            detection["license_expression"]
+            for detection in license_detections
+        ]
+
+        resource.detected_license_expression = combine_expressions(
+            expressions=license_expressions,
+            relation='AND',
+            unique=True,
+        )
+
+        resource.detected_license_expression_spdx = str(build_spdx_license_expression(
+            license_expression=resource.detected_license_expression,
+            licensing=get_cache().licensing,
+        ))
+
+        codebase.save_resource(resource)
+        yield resource
+
+
 def add_license_from_sibling_file(resource, codebase, no_licenses):
 
     if TRACE:

src/packagedcode/plugin_package.py

@@ -27,6 +27,7 @@
 from licensedcode.detection import DetectionRule
 from packagedcode import get_package_handler
 from packagedcode.licensing import add_referenced_license_matches_for_package
+from packagedcode.licensing import add_referenced_license_detection_from_package
 from packagedcode.licensing import add_license_from_sibling_file
 from packagedcode.licensing import get_license_detection_mappings
 from packagedcode.licensing import get_license_expression_from_detection_mappings
@@ -162,6 +163,8 @@ def process_codebase(self, codebase, strip_root=False, **kwargs):
         """
         no_licenses = False
 
+        # These steps add proper license detections to package_data and hence
+        # this is performed before top level packages creation
         for resource in codebase.walk(topdown=False):
             if not hasattr(resource, 'license_detections'):
                 no_licenses=True
@@ -190,6 +193,15 @@ def process_codebase(self, codebase, strip_root=False, **kwargs):
         # Create codebase-level packages and dependencies
         create_package_and_deps(codebase, strip_root=strip_root, **kwargs)
 
+        if not no_licenses:
+            # This step is dependent on top level packages 
+            for resource in codebase.walk(topdown=False):
+                # If there is a unknown reference to a package we add the license
+                # from the package license detection
+                modified = list(add_referenced_license_detection_from_package(resource, codebase))
+                if TRACE and modified:
+                    logger_debug(f'packagedcode: process_codebase: add_referenced_license_matches_from_package: modified: {modified}')
+
 
 def add_license_from_file(resource, codebase, no_licenses):
     """