Add ASYNCIFY+JSPI support to MINIMAL_RUNTIME build mode. #22501
Conversation
| @@ -23,8 +23,15 @@ function assert(condition, text) { | |||
| } | |||
| #endif | |||
|
|
|||
| #if ASYNCIFY == 1 // ASYNCIFY-mode requires checking ABORT variable to avoid operating if code has aborted during an unwind | |||
| var ABORT = 0; | |||
There was a problem hiding this comment.
It seems like this relates to ASYNCIFY == 1 but the PR title talks about JSPI. Does this change cover both?
There was a problem hiding this comment.
Err, yes. Updated the title to clarify that this PR addresses both ASYNCIFY and JSPI.
| @@ -23,8 +23,15 @@ function assert(condition, text) { | |||
| } | |||
| #endif | |||
|
|
|||
| #if ASYNCIFY == 1 // ASYNCIFY-mode requires checking ABORT variable to avoid operating if code has aborted during an unwind | |||
There was a problem hiding this comment.
Is this a debug feature or do we need to add this new variable even in release builds?
I worry that this makes the feature matrix even more complicated and there could be other places in the codebase that deal with ABORT that need to be changes now that ABORT does exist undef MINIMAL_RUNTIME but only with ASYNCIFY is enabled.
There was a problem hiding this comment.
I added that variable to help places like this and this to compile.
There seem to be places that call to it in library_async.js: https://github.com/emscripten-core/emscripten/blob/main/src/library_async.js#L18 so looks like in ASYNCIFY==1 mode we cannot avoid adding ABORT in the build. In ASYNCIFY==2 mode, maybe we could.
Overall, I think the support would be ok to be expanded incrementally via proof by test cases.
juj
changed the title
No description provided.