Merge pull request #20484 from kaspersv/kaspersv/future-proof-java-discarding

Overlay: Future-proof Java config & XML discard predicates

Author: kaspersv

java/ql/lib/semmle/code/java/Overlay.qll

@@ -18,7 +18,7 @@ predicate isOverlay() { databaseMetadata("isOverlay", "true") }
 overlay[local]
 string getRawFile(@locatable el) {
   exists(@location loc, @file file |
-    hasLocation(el, loc) and
+    (hasLocation(el, loc) or xmllocations(el, loc)) and
     locations_default(loc, file, _, _, _, _) and
     files(file, result)
   )
@@ -73,40 +73,22 @@ private predicate discardReferableLocatable(@locatable el) {
   )
 }
 
+/** Gets the raw file for a configLocatable. */
 overlay[local]
-private predicate baseConfigLocatable(@configLocatable l) { not isOverlay() and exists(l) }
-
-overlay[local]
-private predicate overlayHasConfigLocatables() {
-  isOverlay() and
-  exists(@configLocatable el)
-}
-
-overlay[discard_entity]
-private predicate discardBaseConfigLocatable(@configLocatable el) {
-  // The properties extractor is currently not incremental, so if
-  // the overlay contains any config locatables, the overlay should
-  // contain a full extraction and all config locatables from base
-  // should be discarded.
-  baseConfigLocatable(el) and overlayHasConfigLocatables()
-}
-
-overlay[local]
-private predicate baseXmlLocatable(@xmllocatable l) {
-  not isOverlay() and not files(l, _) and not xmlNs(l, _, _, _)
+private string getRawFileForConfig(@configLocatable el) {
+  exists(@location loc, @file file |
+    configLocations(el, loc) and
+    locations_default(loc, file, _, _, _, _) and
+    files(file, result)
+  )
 }
 
 overlay[local]
-private predicate overlayHasXmlLocatable() {
-  isOverlay() and
-  exists(@xmllocatable l | not files(l, _) and not xmlNs(l, _, _, _))
+private string baseConfigLocatable(@configLocatable el) {
+  not isOverlay() and result = getRawFileForConfig(el)
 }
 
 overlay[discard_entity]
-private predicate discardBaseXmlLocatable(@xmllocatable el) {
-  // The XML extractor is currently not incremental, so if
-  // the overlay contains any XML locatables, the overlay should
-  // contain a full extraction and all XML locatables from base
-  // should be discarded.
-  baseXmlLocatable(el) and overlayHasXmlLocatable()
+private predicate discardBaseConfigLocatable(@configLocatable el) {
+  overlayChangedFiles(baseConfigLocatable(el))
 }

java/ql/lib/semmle/code/xml/XML.qll

@@ -6,6 +6,7 @@ module;
 
 import semmle.files.FileSystem
 private import codeql.xml.Xml
+private import semmle.code.java.Overlay
 
 private module Input implements InputSig<File, Location> {
   class XmlLocatableBase = @xmllocatable or @xmlnamespaceable;
@@ -69,3 +70,13 @@ private module Input implements InputSig<File, Location> {
 }
 
 import Make<File, Location, Input>
+
+private class DiscardableXmlAttribute extends DiscardableLocatable, @xmlattribute { }
+
+private class DiscardableXmlElement extends DiscardableLocatable, @xmlelement { }
+
+private class DiscardableXmlComment extends DiscardableLocatable, @xmlcomment { }
+
+private class DiscardableXmlCharacters extends DiscardableLocatable, @xmlcharacters { }
+
+private class DiscardableXmlDtd extends DiscardableLocatable, @xmldtd { }