Stream: git-wasmtime

Topic: wasmtime / PR #13154 Refactor reference types in the C/C+...

Wasmtime GitHub notifications bot (Apr 20 2026 at 22:01):

alexcrichton opened PR #13154 from alexcrichton:refactor-gc-c-api to bytecodealliance:main:

In updating wasmtime-py to Wasmtime 44 I noticed a few discrepancies in the C API, and then C++ API, in addition to what was found in #13149. Notably we had both wasmtime_exn_t/Exn and
wasmtime_exnref_t/ExnRef. In sorting this out with some deeper fixes in #13149 I got stuck again in the rat's nest of circular definitions of classes in C++. In the end I gave up trying to incrementally improve things and decided to go with a larger refactoring. The goal here is to more cleanly separate out responsibilities and improve how circular definitions in C++ are handled:

• All reference types now live in their own file. Reference types aren't in a mix of val.h or gc.h or exn.h, but instead each type lives in its own file. This means that gc.h is now gone and is replaced by anyref.h, externref.h, exnref.h, eqref.h, ...

• C++ class definitions are split into separate headers from their main files. For example there's now wasmtime/_anyref_class.hh instead of just wasmtime/anyref.hh. This new method of resolving circular dependencies has some nice properties where headers that depend on the AnyRef type, for example, include _anyref_class.h. The anyref.h header then is able to include all other headers necessary for defining methods. Notably this means that methods now always live in the header of the type as opposed to sprinkled about in various locations. Note that this is applied to preexisting headers like func.hh and store.hh out of necessity, too.

• Reference types in the C++ now share core methods via a macro to avoid duplicating code and to ensure they have a uniform API.

• Some #defines for WASMTIME_FEATURE_GC guards were adjusted in a few locations. For example wasmtime_valraw_t no longer has GC fields when the GC proposal is disabled.

• Reorganization was reflected on the Rust side as well. This means the previous ref.rs is now broken up into files such as anyref.rs. Some logic of core functions was also deduplicated within the ref_wrapper! macro. The old definitions of wasmtime_exn_t and wasmtime_exnref_t were also both merged.

Overall this PR contains quite a lot of code movement and things were shuffled around, as well as adjustments to the exnref C API (removing duplication). Overall though the API is largely as it was before, although for the C++ API if individual headers are being included some new ones may need to be included to ensure all methods are defined.

My hope is that the end result here is a bit more maintainable, types are easier to modify/extend, there's no longer any soups to deal with in C++, and everything should have a uniform API.

<!--
Please make sure you include the following information:

• If this work has been discussed elsewhere, please include a link to that
  conversation. If it was discussed in an issue, just mention "issue #...".

• Explain why this change is needed. If the details are in an issue already,
  this can be brief.

Our development process is documented in the Wasmtime book:
https://docs.wasmtime.dev/contributing-development-process.html

Please ensure all communication follows the code of conduct:
https://github.com/bytecodealliance/wasmtime/blob/main/CODE_OF_CONDUCT.md
-->

Wasmtime GitHub notifications bot (Apr 20 2026 at 22:01):

alexcrichton requested dicej for a review on PR #13154.

Wasmtime GitHub notifications bot (Apr 20 2026 at 22:01):

alexcrichton requested wasmtime-core-reviewers for a review on PR #13154.

Wasmtime GitHub notifications bot (Apr 20 2026 at 22:16):

alexcrichton updated PR #13154.

Wasmtime GitHub notifications bot (Apr 20 2026 at 23:11):

alexcrichton updated PR #13154.

Wasmtime GitHub notifications bot (Apr 20 2026 at 23:14):

alexcrichton updated PR #13154.

Wasmtime GitHub notifications bot (Apr 21 2026 at 03:49):

github-actions[bot] added the label wasmtime:c-api on PR #13154.

Wasmtime GitHub notifications bot (Apr 21 2026 at 13:31):

fitzgen submitted PR review:

LGTM

I think in the few other C++ codebases I've touched, the _foo.hh would be foo.hh and the existing foo.hh would be foo-inlines.hh but I don't really care what color we paint the bikeshed.

Wasmtime GitHub notifications bot (Apr 21 2026 at 13:31):

fitzgen created PR review comment:

// ============================================================================
// wasmtime_array_type_t

?

Wasmtime GitHub notifications bot (Apr 21 2026 at 13:31):

fitzgen created PR review comment:

Awkward indent, I think it should be something like

        )?
    }) => {

to match the opening

Wasmtime GitHub notifications bot (Apr 21 2026 at 13:31):

fitzgen created PR review comment:

 * APIs for interacting with WebAssembly GC `structref` type in Wasmtime.

Wasmtime GitHub notifications bot (Apr 21 2026 at 14:42):

alexcrichton updated PR #13154.

Wasmtime GitHub notifications bot (Apr 21 2026 at 14:43):

alexcrichton commented on PR #13154:

Oh nice! I'll be honest in that I don't really have much experience in C++ codebases. I'm going to flag this for merge mostly out of inertia, but I'm also happy to adjust to that pattern next time I'm in these parts.

Wasmtime GitHub notifications bot (Apr 21 2026 at 14:43):

alexcrichton has enabled auto merge for PR #13154.

Wasmtime GitHub notifications bot (Apr 21 2026 at 15:16):

alexcrichton added PR #13154 Refactor reference types in the C/C++ API to the merge queue.

Wasmtime GitHub notifications bot (Apr 21 2026 at 15:41):

alexcrichton merged PR #13154.

Wasmtime GitHub notifications bot (Apr 21 2026 at 15:41):

alexcrichton removed PR #13154 Refactor reference types in the C/C++ API from the merge queue.

Last updated: May 03 2026 at 22:13 UTC