Rust: Update legacy MaD models 4

[geoffw0]

Final update of legacy MaD models to the new model format (continues from #19946 and should be independent of that).

There's a lot of guesswork involved in this last set of changes - i.e. cases where I couldn't find a real getCanonicalPath. There should be tests for most of them though so we will find out if we're way off.

[geoffw0]

I think it will be prudent to get the first two/three PRs merged before attempting to address the test failures here.

[aibaars]

The test failures look libc related. It looks like the fn free() in libc doesn't have a result for ItemNode::getCanonicalPath() .

The relevant bits of code are

• https://github.com/rust-lang/libc/blob/8e6f36c6ba5c91c150f2d79a06ffdc3d0b83a0f6/src/lib.rs#L75-L83
• https://github.com/rust-lang/libc/blob/8e6f36c6ba5c91c150f2d79a06ffdc3d0b83a0f6/src/unix/mod.rs#L682

I would expect the canonical path to be either libc::unix::free or libc::free.

This should address the underlying issue #19988

[geoffw0]

I've brought in the fix for extern blocks from main, fixed the libc models, and fixed a couple of typos in other models. There are still a few regressions, but I feel they're at a level where we can merge this and come back to them later.

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

[aibaars]

I think all the MISSING annotations can be fixed by updating the models to match the actual canonical paths. The following queries can come in handy to investigate what the computed canonical paths are:

query predicate resolvedPaths2(CallExprBase e, string path, Addressable target) {
  toBeTested(e) and
  target = e.getStaticTarget() and
  path = concat(target.getCanonicalPath())
}

query predicate resolvedPaths3( Addressable target, string path) { path = target.getCanonicalPath() }

[aibaars]

This looks good to me!

[geoffw0]

The following queries can come in handy to investigate what the computed canonical paths are

Yes I've been doing something similar. The issue being that quite a lot of canonical paths were missing (the situation is improving quite fast) so I had to fill in the gaps myself, and I appear to have made a few typos in that process.

[geoffw0]

DCA LGTM. This is ready for approval I think.