PeerID Derivation Is Broken

[bonedaddy]

PeerID derivation is currently broken. It works sometimes, but other times doesn't work and returns bad data. I think there's some kind of memory issue, but am uncertain.

Example:

5UQDbFGV9gpiaEH <-- bad

QmfGFNuk8rk9oQU8iX572JgtExBnwsAmSzDhSwPRdpRLGG <-- good

The error appears to come from mh_new function call

[bonedaddy]

https://tls.mbed.org/discussions/bug-report-issues/hmac-sha256-incorrect-signature
https://forums.mbed.com/c/mbed-tls/bug-report-issues/22

[bonedaddy]

opened up a report on the armbed fforum https://forums.mbed.com/t/sha256-output-is-not-always-32-bytes-about-25-of-the-time/9343

[bonedaddy]

I believe i've found the source of the problem. The library we use for base58 encoding does not properly handle input data that as a 0 in it. It seems like if the input data has a 0 in it, then the base58 library interprets that as the end of the input data.

I can think of a few work arounds:

• hex encode the sha256 hash before converting to peerid
• switch from base58 to base32, or base64