Merge pull request #15936 from yoff/python/test-conflicting-summaries

Python: No `fieldFlowBranchLimit` for `SummarizedCallable`s

python/ql/lib/semmle/python/dataflow/new/internal/DataFlowImplSpecific.qll

@@ -22,4 +22,6 @@ module PythonDataFlow implements InputSig<Python::Location> {
   predicate neverSkipInPathGraph = Private::neverSkipInPathGraph/1;
 
   Node exprNode(DataFlowExpr e) { result = Public::exprNode(e) }
+
+  predicate ignoreFieldFlowBranchLimit(DataFlowCallable c) { exists(c.asLibraryCallable()) }
 }

python/ql/test/experimental/dataflow/summaries/InlineTaintTest.expected

@@ -0,0 +1,4 @@
+argumentToEnsureNotTaintedNotMarkedAsSpurious
+untaintedArgumentToEnsureTaintedNotMarkedAsMissing
+testFailures
+failures

python/ql/test/experimental/dataflow/summaries/InlineTaintTest.ql

@@ -0,0 +1,4 @@
+import python
+import experimental.meta.InlineTaintTest
+import MakeInlineTaintTest<TestTaintTrackingConfig>
+import TestSummaries

python/ql/test/experimental/dataflow/summaries/TestSummaries.qll

@@ -136,3 +136,108 @@ private class SummarizedCallableJsonLoads extends SummarizedCallable {
     preservesValue = true
   }
 }
+
+// Repeated summaries
+private class SummarizedCallableWithSubpath extends SummarizedCallable {
+  SummarizedCallableWithSubpath() { this = "extracted_package.functions.with_subpath" }
+
+  override DataFlow::CallCfgNode getACall() {
+    result =
+      API::moduleImport("extracted_package")
+          .getMember("functions")
+          .getMember("with_subpath")
+          .getACall()
+  }
+
+  override DataFlow::ArgumentNode getACallback() {
+    result =
+      API::moduleImport("extracted_package")
+          .getMember("functions")
+          .getMember("with_subpath")
+          .getAValueReachableFromSource()
+  }
+
+  override predicate propagatesFlow(string input, string output, boolean preservesValue) {
+    input = "Argument[0]" and
+    output = "ReturnValue" and
+    preservesValue = false
+  }
+}
+
+private class SummarizedCallableWithSubpathAgain extends SummarizedCallable {
+  SummarizedCallableWithSubpathAgain() { this = "extracted_package.functions.with_subpathII" }
+
+  override DataFlow::CallCfgNode getACall() {
+    result =
+      API::moduleImport("extracted_package")
+          .getMember("functions")
+          .getMember("with_subpath")
+          .getACall()
+  }
+
+  override DataFlow::ArgumentNode getACallback() {
+    result =
+      API::moduleImport("extracted_package")
+          .getMember("functions")
+          .getMember("with_subpath")
+          .getAValueReachableFromSource()
+  }
+
+  override predicate propagatesFlow(string input, string output, boolean preservesValue) {
+    input = "Argument[0]" and
+    output = "ReturnValue.Attribute[pattern]" and
+    preservesValue = true
+  }
+}
+
+private class SummarizedCallableWithoutSubpath extends SummarizedCallable {
+  SummarizedCallableWithoutSubpath() { this = "extracted_package.functions.without_subpath" }
+
+  override DataFlow::CallCfgNode getACall() {
+    result =
+      API::moduleImport("extracted_package")
+          .getMember("functions")
+          .getMember("without_subpath")
+          .getACall()
+  }
+
+  override DataFlow::ArgumentNode getACallback() {
+    result =
+      API::moduleImport("extracted_package")
+          .getMember("functions")
+          .getMember("without_subpath")
+          .getAValueReachableFromSource()
+  }
+
+  override predicate propagatesFlow(string input, string output, boolean preservesValue) {
+    input = "Argument[0]" and
+    output = "ReturnValue" and
+    preservesValue = false
+  }
+}
+
+private class SummarizedCallableWithoutSubpathAgain extends SummarizedCallable {
+  SummarizedCallableWithoutSubpathAgain() { this = "extracted_package.functions.without_subpathII" }
+
+  override DataFlow::CallCfgNode getACall() {
+    result =
+      API::moduleImport("extracted_package")
+          .getMember("functions")
+          .getMember("without_subpath")
+          .getACall()
+  }
+
+  override DataFlow::ArgumentNode getACallback() {
+    result =
+      API::moduleImport("extracted_package")
+          .getMember("functions")
+          .getMember("without_subpath")
+          .getAValueReachableFromSource()
+  }
+
+  override predicate propagatesFlow(string input, string output, boolean preservesValue) {
+    input = "Argument[0]" and
+    output = "ReturnValue.Attribute[pattern]" and
+    preservesValue = true
+  }
+}

python/ql/test/experimental/dataflow/summaries/conflicting_summaries.py

@@ -0,0 +1,18 @@
+# Bad interaction of two summaries for the same function
+ts = TAINTED_STRING
+
+from extracted_package.functions import with_subpath, without_subpath
+
+# For the function `with_subpath`, flow from the first argument to the return value
+# can be concluded from its definition. This seems to discard all summaries, including
+# the one with flow to `ReturnValue.Attribute[pattern]`.
+ensure_tainted(
+    with_subpath(ts).pattern,  # $ tainted
+    with_subpath(ts),  # $ tainted
+    with_subpath(ts),  # $ tainted
+)
+ensure_tainted(
+    without_subpath(ts).pattern,  # $ tainted
+    without_subpath(ts),  # $ tainted
+    without_subpath(ts),  # $ tainted
+)

python/ql/test/experimental/dataflow/summaries/extracted_package/functions.py

@@ -0,0 +1,5 @@
+def with_subpath(x):
+    return x
+
+def without_subpath(x):
+    pass