Give serialization derives a pair of eyes #127
Conversation
This is a functionality that isn't being utilized. We're relying on `rkyv` for (de-)serialization and storage-at-rest purposes. Users are encouraged to store rkyv payloads inside their serde containers if they need to do so.
Admittedly Rkyv{Des,S}erialize doesn't really read much better than
`rkyv::*` and it is made clearer that it is an external derive as per
nearcore#6596.
|
A lot of error handling in wasmer can be summarized as “a lot of strings”, so it isn't inconceivable to me that there wouldn't be a separate error variant for this. Another reason I have to believe this is probably not the case is that there isn't actually anywhere in the code that utilizes serde to produce encodings. Users AFAIU were expected to do so externally (which makes sense). |
|
I am deeply offended by this PR! It is preposterous that it was merged! It should say ‘quartet of eyes’! ;) |
|
@mina86 I added you review by accident, but, given that I've already distracted you, the context for this mistake is that we are debating whether doing similar treatment to nearcore would be a good idea, and I though you might have opinions here. See near/nearcore#6596 for ctx ::) |
Best viewed commit-by-commit.
This is in a way a test drive for near/nearcore#6596