ComputeEcdsaSecp256k1Key is never used #1087
Assignees
Labels
bug
Something isn't working
Comments
github-merge-queue bot
pushed a commit
that referenced
this issue
Oct 11, 2023
Today I didn't get as much done -- had some errands and a doctor appointment -- but my goal was to get through the residual bits of host.rs and I did that at least. Along the way I removed the `allows` that enabled un-warned dead code and unused variables, and found some stuff: - Two missing error checks on linear-memory "unpack" vector lengths. These don't represent a _major_ vulnerability, but they do allow users to do an unpack to a mismatched vector length, which could hide a bug. A footgun at least. Fixed. - A dead cost type, based on a previous iteration of the secp256k1 interface. Filed #1087 to track removing it. - Some diagnositcs weren't using the names we so meticulously pass down to them. Just throwing them out. Fixed. - There was a bunch of dead code and over-public interfaces. Fixed.
|
Xdr change has been merged. Env change will be made with #1105. |
github-merge-queue bot
pushed a commit
that referenced
this issue
Oct 12, 2023
### What - Update to the latest XDR, consolidating a few memory-related `ContractCostType` -- resolves #1020 - Resolve #1087 - Move some no longer used cost type calibrations (VecEntry, MapEntry) to an `experimental` directory, they are currently not used anywhere but would be useful for experimental purpose. Will do a followup to make them usable. - Fixed a bug in `memory_grow` function, where we were checking the limit against the wrong input - Add test helpers to make wasm memory alloc accessible - Various test fixes, clearifications ### Why [TODO: Why this change is being made. Include any context required to understand the why.] ### Known limitations Follow ups: - Refactor the analytical models - Make `experimental` directory usable - Recalibrate numbers on an x86 machine and update the parameters - Write a more complex test than the currently `complex`, use that for metering benchmark --------- Co-authored-by: Graydon Hoare <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
The interface we settled on for Secp256k1 is the "key recovery" interface which doesn't involve a user ever converting a bytes object directly to a public key. This cost is therefore never used (and the codepaths that do it are only used to calibrate the cost, never actually do it).
It should probably be removed from the XDR cost types.
The text was updated successfully, but these errors were encountered: