C++: Get rid of an ugly workaround in dataflow

[MathiasVP]

For a long time we've had an ugly hack in the C/C++ dataflow library related to postfix operator uses. This hack comes from the fact that we want dataflow in cases like:

int x = source();
sink(x++);

but not in:

int x = source();
x++;
sink(x);

The hack got this by modifying the location of the read attached to x++. However, it's both ugly and error prone (and, in fact, for some range analysis work I'm doing this was causing infinite looping in the shared range analysis library).

This PR gets rid of this hack and instead properly models this in our SSA library. So for something like:

int x = source();
sink(x++);

we create a new SSA source variable, definition and use such that the above program is more faithfully represented as:

int x = source();
int x_tmp = x;
x++;
sink(x_tmp);

this ensures that we get the correct dataflow behavior.

Commit-by-commit review recommended.

• bd09aad adds loads of tests
• e0a7889 reverts the hack. You can see how this makes the tests regress.
• 3d445be and db3f22a performs small refactorings that shouldn't have any effect on any results, but will make the actual change much easier.
• Finally, 445cca1 implements the strategy outlines above, and
• 28fec0c shows that the regressions are fixed.

I've tested that the isUseAfterPostfixCrement0 predicate indeed correspond exactly to the uses of postfix increments and decrements in lots of databases (by checking whether all the StoreInstructions for which isUseAfterPostfixCrement0(store, _) holds come from here, and vice versa).

[Copilot]

Pull request overview

Refactors the C/C++ IR dataflow SSA implementation to correctly model postfix increment/decrement by introducing an explicit “saved value” SSA variable, eliminating the previous location-based workaround and restoring correct dataflow behavior.

Changes:

• Added new dataflow tests covering postfix/prefix crements, compound assignments, and conditional-expression cases.
• Updated expected AST/IR flow outputs for the new tests.
• Updated SSA implementation to represent postfix crements via dedicated “saved” source variables/defs/uses, replacing the prior hack.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
cpp/ql/test/library-tests/dataflow/dataflow-tests/test.cpp	Adds new test cases validating desired postfix crement flow behavior and related edge cases.
cpp/ql/test/library-tests/dataflow/dataflow-tests/test-source-sink.expected	Updates expected AST/IR flow results corresponding to the new tests.
cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/SsaImpl.qll	Implements the new SSA modeling for postfix crements by adding saved-value source variables and wiring corresponding defs/uses.

Comments suppressed due to low confidence (2)

cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/SsaImpl.qll:50

• Same grammar nit in the doc comment here: "generated by an postfix" should be "generated by a postfix".

   * Holds if `store` is the `StoreInstruction` generated by an postfix

cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/SsaImpl.qll:697

• This call site uses hasBaseSourceVariableAndIndirectrion; if you fix the typo in the predicate name, make sure this reference is updated to match.

      this.hasBaseSourceVariableAndIndirectrion(v, indirection)

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.