From 36914dcb528db9e79fbb8658c20131e58038de83 Mon Sep 17 00:00:00 2001
From: Kristen Newbury <knewbury01@github.com>
Date: Fri, 18 Jul 2025 13:15:50 -0400
Subject: [PATCH] Add taint debug notes

---
 codeql-sqlite/TaintFlowDebugging.md | 65 +++++++++++++++++++++++++++++
 1 file changed, 65 insertions(+)
 create mode 100644 codeql-sqlite/TaintFlowDebugging.md

diff --git a/codeql-sqlite/TaintFlowDebugging.md b/codeql-sqlite/TaintFlowDebugging.md
new file mode 100644
index 0000000..cc26d48
--- /dev/null
+++ b/codeql-sqlite/TaintFlowDebugging.md
@@ -0,0 +1,65 @@
+# Debugging with partial flow
+
+When there is a missing node on a path that you are trying to define there are 2 ways to figure out what the missing node is:
+
+1) Using the predicate `any()` in place of a sink. [QL documentation](https://codeql.github.com/docs/writing-codeql-queries/debugging-data-flow-queries-using-partial-flow/#partial-flow) explains the details of why this is not too performat, but it works fine when an application is small or you have a confident idea of what the missing piece is and really only need a small quick check to help out.
+
+2) [Partial flow](https://codeql.github.com/docs/writing-codeql-queries/debugging-data-flow-queries-using-partial-flow/#debugging-data-flow-queries-using-partial-flow) dataflow configuration is the preferred, more thorough solution for debugging dataflow. This configuration can allow you to see the partial pathes either forward or backward through the application and answer a question of "how far does the current path get?". This is extremely useful for cases where a node is missing in the dataflow graph, and an additional taint step is required to model the full problem.
+
+
+The QL below demonstrates what partial debugging would look like, on [this Java SqlInjection sample](https://github.com/advanced-security/codeql-workshops-staging/tree/master/java/codeql-dataflow-sql-injection).
+
+```
+/**
+ * @name introduction workshop
+ * @description Sample SQL Injection problem
+ * @id test
+ * @kind path-problem
+ * @problem.severity warning
+ */
+
+import java
+
+class ReadLineSource extends Source {
+  ReadLineSource() { this.getMethod().hasQualifiedName("java.io", "Console", "readLine") }
+}
+
+abstract class Source extends MethodCall { }
+
+class Sink extends MethodCall {
+  Sink() { this.getMethod().hasQualifiedName("java.sql", "Statement", "executeUpdate") }
+}
+
+import semmle.code.java.dataflow.TaintTracking
+
+module MyFlowConfiguration implements DataFlow::ConfigSig {
+  predicate isSource(DataFlow::Node source) { source.asExpr() instanceof Source }
+
+  predicate isSink(DataFlow::Node sink) {
+    exists(Sink sink2 | sink.asExpr() = sink2.getArgument(_))
+    //previous debug technique shown , not ideal though
+    //any()
+  }
+
+  //this is the necessary flow step to close the gap
+  //   predicate isAdditionalFlowStep(DataFlow::Node inNode, DataFlow::Node outNode) {
+  //     exists(MethodCall mc |
+  //       outNode.asExpr() = mc and
+  //       mc.getMethod().hasQualifiedName("java.lang", "String", "format") and
+  //       inNode.asExpr() = mc.getAnArgument()
+  //     )
+  //   }
+}
+
+int explorationLimit() { result = 100 }
+
+module MyFlow = DataFlow::Global<MyFlowConfiguration>;
+
+module MyPartialFlow = MyFlow::FlowExplorationFwd<explorationLimit/0>;
+
+import MyPartialFlow::PartialPathGraph
+
+from MyPartialFlow::PartialPathNode start, MyPartialFlow::PartialPathNode end
+where MyPartialFlow::partialFlow(start, end, _)
+select end, start, end, "Sql injection from $@", start, "here"
+```
\ No newline at end of file