Jake Hillion

No longer failing with codegen-v2, but the results are weird: in the top level case (`IncompleteStruct*`) you get an entry for the `IncompleteStruct`, but with other cases like `std::unique_ptr` you...

Sorry for not picking up on that. The approach taken with RocksDB seems good (if !find_package then fetchcontent). Could you also create an issue for changing this to `FIND_PACKAGE_ARGS` when...

To add, this was also missing in codegenv1 - we've never had good support for jit logging in containers.

I'm wondering if this change makes total sense - we're taking in a function type which is explicitly unsized, and stating that it's a sized type `StubbedPointer`. I wonder if...

Thanks for trying! I'll go through the bits you mentioned then say what I think will be the best solutions. > * the rocksdb version you reference doesn't support libstdc++...

Can be fixed with `expect_json_v2` in the meantime

Closing this out, OIL v2 runs on this approach and OID support may come in the future as time permits, this issue isn't needed to track it.

How often is it that we use the same data segment twice anymore? I wonder if instead of complicating the logic we could always clean up the data segment when...

- `OidIntegration.pointers_void` - `OidIntegration.pointers_void_no_follow` - `OidIntegration.pointers_void_null` Are special cases of this. As oid codegen sees them as the underlying type (`void` when it should be `void*`) it hits: https://github.com/facebookexperimental/object-introspection/blob/1c3ee5bf6b79676bce4155284f3723de42b2227e/src/OICodeGen.cpp#L1025-L1028

We could probably do this with a simple optional text field in the config. If we always set the defaults then follow up with e.g.: ``` opts.cacheDirPath = "..."; opts.generateJitDebugInfo...