Innokentii Sennovskii

Dangerous lack of subgroup check for G2 groups in bls

3

### What is wrong? G2 point decompression function goes through all the regular checks same as for G1 (checks that coordinates are in field and that the point is on...

Could you tell me why this has only 1 path?

4

``` python from symbolic.args import * @symbolic(a=0xdeaddeaddeaddead,b=0xbeefbeefbeefbeef) def expandKey(a,b): i=0; passkeyn=[a,b] expandedkey=[] while i>0x30)|(((v1>>0x20)&0xffff)0x10)&0xffff)0x20)&0xffff)0x10)&0xffff)0x30)|(((v5>>0x20)&0xffff)0x10)&0xffff) 39)0x30)|(((v6>>32)&0xffff)0x10)&0xffff) 39)

Randomness in barretenberg is taken from random_device which is extremely slow (syscall per 4 bytes). It needs to be replaced at some point at least for the server usecase. The...

IPA needs docs in-code

Create_sort_constraint_with edges goes through variables in quadruplets. However, for the last row it takes the last 4 variables in the list. There might be a correctness issue with this approach,since...

Currently if we submit the point at infinity to the transcript, it will not detect the issue and will store a malformed representation of the point, leading to errors down...

Some affine element tests are defined just for one curve. Absolutely no reason for that

Affine and group elements in barretenberg have a special point at infinity member which can't just be encoded as fields, which is what we currently do with msgpack. Changing it...

Translator Composer uses a random challenge that is derived from Fiat-Shamir, while circuit builder fuzzer uses a configurable challenge. We need to reconcile the two, so they are the same...

We can use tags to build parts of docs in parallel (NUM_PROC_THREADS doesn't change much). Right now it takes 16 seconds to build docs, which makes iterating painful https://www.doxygen.nl/manual/external.html