Add cxx_virtual_method instruction.

[pschuh]

This also contains the code to trigger forming cxx_virtual_method instructions but the this pointer adjustment is not hooked up yet.

Companion pr: apple/swift-clang#349

[jrose-apple]

Hmm. I'm sad that we need cxx_virtual_method at all because it does not bode well for adding other importers in the future. I'll take a closer look soon.

[pschuh]

Hi @jrose-apple, the context for needing cxx_virtual_method is here: https://forums.swift.org/t/c-interop-virtual-function-calls/27567/3 in case you missed it.

[jrose-apple]

I guess if we have a C++ abstraction pattern we can have a cxx_virtual_method instruction.

[Azoy]

I'm just a simple bystander, but could you add some SILGen tests as well?

[jrose-apple]

I wonder if there's a better way to have exactly two results than this. My SIL isn't strong enough. @gottesmm?

[jrose-apple]

@rjmccall or @jckarter should probably verify this rather than me.

[jrose-apple]

This seems important…

[jrose-apple]

When you do get to it, we'll need some way to indicate that the adjusted "this" pointer depends on the lifetime of the original object. I'm not sure if there's a way to do that other than the mark_dependence instruction, but there ought to be.

[gottesmm]

My suggestion (going back to what I was saying above) is that the method should take self either forwarding or guaranteed.

[gottesmm]

Before you land this, I want to take another look through. But some initial comments.

[gottesmm]

This is the wrong semantics. The right semantics IMO are that this should be either forwarding or guaranteed.

[gottesmm]

My suggestion (going back to what I was saying above) is that the method should take self either forwarding or guaranteed.

[shahmishal]

Please update the base branch to main by Oct 5th otherwise the pull request will be closed automatically.

• How to change the base branch: (Link)
• More detail about the branch update: (Link)