alexcrichton commented on issue #6509:
I'll note that I had a number of false starts in attempting to solve this issue. First I thought that this only needed a push/pop around calls to
cabi_realloc
in the component model, but that wasn't correct because that only works for import-based calls torealloc
, not export-based calls. I then wrote a test to expose the flaw with that strategy.Next I thought that it would be possible to not actually maintain the entire linked list of activations and instead have a fiber have its own linked list which replaces the current thread's linked list when it resumes and swaps it back out when it suspends. This however does not work because a complete list of activations is required to correctly perform GC, so I wrote a test to expose that bug.
Finally I settled on this solution, which is at least sufficient for all the above scenarios. I'm not 100% sure it's the end-all-be-all, so extra care in review would be appreciated!
github-actions[bot] commented on issue #6509:
Subscribe to Label Action
cc @peterhuene
<details>
This issue or pull request has been labeled: "wasmtime:api"Thus the following users have been cc'd because of the following labels:
- peterhuene: wasmtime:api
To subscribe or unsubscribe from this label, edit the <code>.github/subscribe-to-label.json</code> configuration file.
Learn more.
</details>
alexcrichton commented on issue #6509:
I definitely agree with the separation of types and re-wording of things, done now :+1:
Last updated: Dec 23 2024 at 13:07 UTC