C++: Expose more information in FormattingFunction and make subclasses private.

[geoffw0]

Follow-up from #4634

• expose information about how the formatted string is output though new predicates in FormattingFunction
• update uses of subclasses Printf, Fprintf etc to use the new predicates instead
• make the subclasses Printf, Fprintf etc private

The tricky part was getting UserDefinedFormattingFunction right (i.e. handling user wrappers to formatting functions). It was already implemented with mutually recursive predicates primitiveVariadicFormatter etc underneath, I believe I've managed to upgrade them and simplify the interface to them without adding much bulk to the implementation.

In the case of AttributeFormattingFunction , we simply do not know what is done with the formatted string, so the new predicates are not implemented.

The test is for a special case, where a user wraps vsprintf or vfprintf in such a way as it doesn't use an output parameter from the caller, and is thus behaving like vprintf. My first attempt to implement the UserDefinedFormattingFunction changes was confused by this case.

[jbj]

Otherwise LGTM

[jbj]

Tests failed:

--- expected
+++ actual
@@ -1,3 +1,4 @@
+WARNING: Predicate variadicFormatter has been deprecated and may be removed in future (variadicFormatter.ql:4,7-24)
 | __vswprintf_l | 1 |
 | _vsnprintf_l | 2 |
 | _vsnprintf_s | 3 |
[6/8] [88/522 comp 2.7s eval 36ms] FAILED(RESULT) /home/runner/work/semmle-code/semmle-code/semmlecode-cpp-tests/library-tests/printf/formattingFunction/variadicFormatter.ql

[geoffw0]

Created https://github.com/github/semmle-code/pull/38334 (internal link) to fix the failing test. Both PRs should be merged at about the same time.

[MathiasVP]

LGTM!