State Tracing #134
State Tracing #134
Conversation
update sqlx from beta to 0.4.0 release
|
Didn't expect a review so fast! There are some problems with the current implementation I wrote up and was going to ask for feedback today State tracing seems to work -- sort of. Replacing WASM code at runtime and tracing itself is working, but Substrate Archive is getting traces for only some parts of the substrate landscape: Haven't made the migrations yet because i'm not sure if there might be something wrong with how i've setup the tracing. So far, it just logs the |
insipx
requested review from
maciejhirsz,
dvdplm and
mattrutherford
and removed request for
dvdplm
| let executor = | ||
| LocalCallExecutor::new(backend.clone(), executor, Box::new(TaskExecutor::new()), Default::default())?; | ||
| let executor = NativeExecutor::<Dispatch>::new( | ||
| WasmExecutionMethod::Interpreted, |
There was a problem hiding this comment.
why not WasmExecutionMethod::Compiled? It's enabled in this PR and I suppose it's faster?!
There was a problem hiding this comment.
I measured this and it's actually not faster. I was surprised by this and it might be worth taking a deeper look. When syncing the chain in normal polkadot using --wasm-execution=Compiled is much much faster.
Co-authored-by: Niklas Adolfsson <[email protected]>
…e-archive into insipx/state-tracing
- Remove clones in Trace database insert - remove frame-benchmarking dependency from polkadot-archive - add new WASM binaries for 8.27 (older ones didn't work) - add comment explaining why we unwrap Arc and Mutex
Continuation of work from #92
Not ready for merge yet because:
In this PR
ActorContext->SystemConfignative_executor_instancein order to have access and change around the HostFunctions available in the executing WASM environment