C++: Make swap member functions data-flow functions

[paldepind]

This PR makes swap member functions DataFlowFunctions instead of TaintFunctions.

This causes some good results in the AST based taint tracking to disappear. We'll have to do some more investigation to figure out why that is the case. I've not updated the tests to expect this regression, so CI will fail.

[paldepind]

This shows that the change is doing what we want for the IR based taint tracking.

[geoffw0]

LGTM, but it looks like you still need to accept changes to some other tests to get CI passing.

[paldepind]

I've figured out why we loose some of the edges in the AST-based taint tracking. It turns out that some cases related to member functions are handled for TaintFunctions but not for DataFlowFunctions. For instance, the disjunct here allows for this to flow to an argument for a TaintFunction, but a similar case does not exist earlier in the predicate for DataFlowFunctions. A similar asymmetry exist in other predicates in that file.

I've quickly tried adding some of these extra cases for DataFlowFunctions and this does fix the missing AST-based flow. But, given that the library is deprecated and subject for removal it doesn't seem like this is worth fixing (a fix that could have other ramifications perhaps). So now that we know the reason I think it's ok to keep the PR as is and accept the lost edges.

[geoffw0]

a fix that could have other ramifications perhaps

Since the library is deprecated we should no longer use it in any of our queries, so there should be no ramifications on results in our query suites and I think no ramifications on performance either. It could well affect users who are still using the deprecated libraries though. We generally expect users to move away from deprecated libraries within 1 year at which point we delete them.

I agree it probably isn't worth fixing, though I'm not strongly against it if you want to.

[paldepind]

Since the library is deprecated we should no longer use it in any of our queries, so there should be no ramifications on results in our query suites and I think no ramifications on performance either. It could well affect users who are still using the deprecated libraries though. We generally expect users to move away from deprecated libraries within 1 year at which point we delete them.

Makes sense.

I agree it probably isn't worth fixing, though I'm not strongly against it if you want to.

Let's keep it as is then 👍

[geoffw0]

The force-push made this a little harder to re-review than it should have been, but LGTM.

[paldepind]

Thanks for the review and sorry about force pushing. I was thinking that as the PR was previously marked as a draft I could get away with it. But it makes sense to not do that given that you'd already reviewed 👍