Closed aulneau closed 1 year ago
ah, applies to btc-signer, too:
Wrong object type for transaction options: [object Object]
It's most likely the case that there are different Object
instances in these environments, which is why the check is failing
I think constructor.name was unreliable 10 years ago. Isn't it still? What about minification? Are you confident the solution would work, did you test it?
We have other checks, like: a instanceof Uint8Array
, which, I assume, would also fail in the scenario.
I have tested it: https://github.com/aulneau/noble-edge-bug/blob/main/src/pages/api/index.ts
Here is a reproduction. I'm not familiar with the reliability of constructor.name or not unfortunately.
@paulmillr anything i can do to help move this on?
I think we can remove obj.constructor === Object
check. Object.prototype.toString.call
is already done. That would make it return different result for ({__proto__: []})
, but if tests pass, I think it could work.
The function is used in only a few places.
@aulneau fixed!
When trying to use some of these hashing functions in the context of a serverless environment, such as vercel or cloudfare, this function
https://github.com/paulmillr/noble-hashes/blob/54c3197fc53bbfa4521b6883eed0e7bd1430beab/src/utils.ts#L163
fails to accept this object
{c:2048,dkLen:64}
as a plain object.I have followed it down to this line:
obj.constructor === Object
never being true. I'd suggest something like this instead:obj.constructor.name === 'Object'