Arvid Norberg
Arvid Norberg
these are some really good finds. I'll work on fixing this as well as attempting to prevent these mistakes to be made in the future
the streamable format of `Offer` was missing test coverage before this change
I suspect that this will make the code run slower, because of the extra indirections. Is there a stronger argument than "generally bad practice"?
we have measured performance degradation with smaller changes in the past. If you can demonstrate that there's no slow-down, I think this is a fine change. I think your use...
on the other hand, given bladebit, perhaps we don't care so much about a performance degradation in chiapos.
this should also be addressed in the `Cargo.toml` file, right? Also, once addressed, we should remove this exception: https://github.com/Chia-Network/clvm_rs/blob/main/.github/workflows/dependency-review.yml#L22C24-L22C43
at least `chia_rs` depends on the buffer protocol API, which is not part of abi3, but is part of all python versions we support. I don't have a strong opinion...
@dependabot rebase
you mean having the allocator contain `PyBuffer` objects, essentially, to keep python buffers alive? I don't think there's a way to expose types with any lifetime annotations. But a wrapper...
in its current form, this patch causes a performance regression. When testing on RPi5: ``` run_program/block-2000 time: [366.63 ns 366.75 ns 366.87 ns] change: [+1.0676% +1.2314% +1.4306%] (p = 0.00...