Hi @tlively. I have motivation to implement this. I hope I don't bother you with some questions. From bird's-eye the lower wasm exceptions to MVP seems doable.

My first idea is to replace the try blocks with just a block with br statement. When a br occurs, I store the exception data somewhere in memory and then process the blocks associated with the catches until exception is consumed.

Something like this:

**FROM**
try 
  throw ...
catch
  ...

**TO**
block
  // store exception data
  br 1

// catch realted to thrown exception
block
  // load exception data

I excuse that it very raw, but I don't want to implement something not working :)

Anyway,

when we up to function level, after each function call we need to check if we processing some exception right now. This means that I need to put some logic after each call/call_indirect instruction. Do you think is it will bloat the binary very fast?
try/delegate instruction, I don't have an idea how to handle it, it breaks the linear execution flow, and makes this simple idea not working. Maybe there is option to not generate them, but again it will bloat binary.
Can we just get implementation from wasm2c? Even more what happen if I do wasm2c -> c2wasm loop, can it work as workaorund?

Uh oh!

Support try-catch in Asyncify / lower wasm exceptions to MVP #5343

Description

Metadata

Metadata

Assignees

Labels

Type

Fields

Projects

Milestone

Relationships

Development

Issue actions