This adds IndirectMustWriteSideEffects for constructor qualifiers. The
introduced sanity failures result from constructor calls without qualifier
operands in the IR
The key insight here is that `HC_FieldCons` and `HC_Array` are
functionally determined by the things that arise in another
recursive call. Lifting them to their own predicate, therefore,
reduces nonlinearity and constrains the join order in a way that
cannot be asymptotically bad -- and, indeed, makes quite a big
difference in practice.
Switching `security.TaintTracking` to use `DefaultTaintTracking` causes
us to lose a result from `UnboundedWrite.ql`, while this commit restores
it:
diff --git a/semmlecode-cpp-tests/DO_NOT_DISTRIBUTE/security-tests/CWE-120/CERT/STR35-C/UnboundedWrite.expected b/semmlecode-cpp-tests/DO_NOT_DISTRIBUTE/security-tests/CWE-120/CERT/STR35-C/UnboundedWrite.expected
index 1eba0e52f0e..d947b33b9d9 100644
--- a/semmlecode-cpp-tests/DO_NOT_DISTRIBUTE/security-tests/CWE-120/CERT/STR35-C/UnboundedWrite.expected
+++ b/semmlecode-cpp-tests/DO_NOT_DISTRIBUTE/security-tests/CWE-120/CERT/STR35-C/UnboundedWrite.expected
@@ -1,2 +1,3 @@
+| main.c:54:7:54:12 | call to strcat | This 'call to strcat' with input from $@ may overflow the destination. | main.c:93:15:93:18 | argv | argv |
| main.c:99:9:99:12 | call to gets | This 'call to gets' with input from $@ may overflow the destination. | main.c:99:9:99:12 | call to gets | call to gets |
| main.c:213:17:213:19 | buf | This 'scanf string argument' with input from $@ may overflow the destination. | main.c:213:17:213:19 | buf | buf |
The user knows that an expression functionally determines its
hashCons value, and that an expression functionally determines
its number of children, but this is not provable from the
definitions, and so not usable by the optimiser. By storing
the result of those known-functional calls in a variable,
rather than repeating the call, we enable better join orders.
The data flow library conflates pointers and their objects in some
places but not others. For example, a member function call `x.f()` will
cause flow from `x` of type `T` to `this` of type `T*` inside `f`. It
might be ideal to avoid that conflation, but that's not realistic
without using the IR.
We've had good experience in the taint tracking library with conflating
pointers and objects, and it improves results for field flow, so perhaps
it's time to try it out for all data flow.
This is for symmetry with `exprNode` etc., and it should be handy for
the same reasons. I found one caller of `asInstruction` that got simpler
by using the new predicate instead.
