Matt Caswell

Possible fix in #19779. Note this only affects master so a work around is to simply compile the latest stable version (3.0.7) instead.

> @levitte can you explain why an atexit handler would be needed in the first place? Back in the 1.0.x days, OpenSSL library init and deinit was a mess. There...

If everything is moved to libctx then I could eventually envisage `OPENSSL_cleanup()` as being synonymous with: ```` OSSL_LIB_CTX_free(OSSL_LIB_CTX_get0_global_default()); ```` > But, aren't we still getting into trouble with the default...

Marked as draft until #19703 is merged.

> Running the fuzzer from https://github.com/openssl/openssl/pull/19743 on top of this, directly generates a memory leak Thanks. Fixup for that pushed.

> After setting up alpn, I get: My fixup wasn't quite right. I've made another attempt. I've also rebased it to pick up the latest version of #19703.

Pushed another couple of fixups. Mostly fixing some cosmetic issues, but also fixing a missing call to `BIO_set_retry_write()`.

> Currently don't see anything obvious why it's not doing anything useful, but I can also not see it actually processing any data. It seems ossl_quic_do_handshake() just says it wants...

Fixup addressing Hugo's comment pushed.

I've now rebased this on top of PR #19734 (QUIC Test Server and Basic Echo Test) instead of #19703, and extended it with an additional commit to supply the required...