Richard Levitte

> Exactly why the locking call crashes, I don't know. Probably because it's not properly initialised in `libcrypto.1.1.dylib`... Yup, that's it. There's a global engine lock, which is initialized as...

Noted. I've seen something similar (not XOF related, though) in another provider I wrote. Maybe the reasons are similar... I'll have a look in a couple of days.

It's a bit interesting, though, that the Github Actions don't fail. I wonder why that is... Why I see on the provider test failure above is that the `size` macopt...

The only solution I found was to add this line in the generated `master.cfg`: ``` python c['buildbotURL'] = "https://buildbot:8010/" ```

I usually use the full path, i.e. `--python=/usr/bin/python3.5`

I like a lot of what I see here and am prepared to give is a try or even help out. Regarding `os` vs `dist`, If you look back at...

@mtrojnar, functions such as `RSA_new_method()` have a `ENGINE_init()` when an engine is provided (i.e. when the `engine` parameter is non-NULL), and they do attach the pointer to the engine in...

Sorry, my comment there was a side note and an answer to a question @mtrojnar asked higher up. I'm currently looking through #107 and trying to understand what's really going...

You know, if I understand things correctly, the _real_ issue here is that OpenSSL's apps/apps.c:setup_engine() calls `ENGINE_set_default()`. Without that call, the methods you set in the engine will only be...

Around some of your questions and assumptions: the ENGINE stuff was never written to support the use of several engines at the same time. It's really a one-engine-per-process deal...