Andy Green

Crash backtraces won't really help with that... if you want to help debug it, running it under linux + valgrind might provide useful information about what the exact code path...

It's less than ideal because the lws api needs to work with OpenSSL if that's what you're using, as well as mbedtls. The two tls libs seem to have fundamentally...

I see, it's basically a fork of boringssl, which we already support. AFAIK it shouldn't be too tricky then. Sure, push the patches somewhere and I will look at them.

You can set the conceal count to `LWS_RETRY_CONCEAL_ALWAYS` which is 0xffff. But the `secs_since_valid_ping` doesn't particularly understand 0xffff (it's a uint16_t). Assuming you mean it is constantly disconnecting under those...

> I guess it might be related to the parameter secs_since_valid_hangup No, as I said >> You can set the conceal count to LWS_RETRY_CONCEAL_ALWAYS which is 0xffff That is related...

Not sure I understand why env vars are "locked" by some other process.... env vars are local to each process, right? Do you mean that the file at the path...

Just some gentle advice if you want to get your patch in, it's better to engage with what the project is saying about it. For your situation it might be...

Thanks... I don't really want to change the behaviour for the normal case that the address is a DNS name, as this patch seems to do. For the vast majority...

``` /home/sai/jobs/85984169/src/ebuild/esp-heltec-wb32/libwebsockets/lib/tls/mbedtls/wrapper/library/ssl_lib.c: In function 'SSL_CTX_new': /home/sai/jobs/85984169/src/ebuild/esp-heltec-wb32/libwebsockets/lib/tls/mbedtls/wrapper/library/ssl_lib.c:208:25: error: 'rngctx' undeclared (first use in this function) 208 | cert = ssl_cert_new(rngctx); ``` ... did you build-test it?

What do you think should happen about that?