andig
ping @marcj is the cgi option still relevant? Doesn't feel like it since we're handling all connection through reactphp?
It should be long enough to cater for startup time of a worker which might be slow?
Check latest react/http release. Remember reading something that might be related.
> I also thought react/http 1.2 would fix this Master allows 1.3 now. Does that help?
Could you provide an example to repro this?
Actually, I feel this is the desired behavior. If client throws without catching we close the slave as we don't know whats wrong. What is the actual problem you're having...
To summarize the discussion sofar: currently- when client throws- PPM returns: HTTP/1.1 502 Bad Gateway Slave returned an invalid HTTP response. Maybe the script has called exit() prematurely? Instead, it...
There are some other issues like https://travis-ci.org/php-pm/php-pm/jobs/617713470#L365, please take a look. I don't see any related to Symfony.
> But I keep having some bugs with return value of execute's command That would need a separate PR, didn't catch that when Symfony 5 was allowed. See https://github.com/php-pm/php-pm/pull/479.
Looking good to me. I imagine a more typical scenario would be to deploy ppm behind an SSL-terminating reverse proxy (Traefik etc). ping @marcj for approval