fatal usually appends the error string. Add 'fatalx' that doesn't. Fix callers and move the prototypes to log.h

it’s the QUERY_STRING decoded if it’s a search-string (i.e. not a key-value pair.) It’s useful for scripts to avoid percent-decoding the querystring in the most common case of a query, because in Gemini querystrings key-value paired are not common. Idea from a discussion with Allen Sobot.

this fixes a possible crash if `client_write' closes the connection, because client_close can end up freeing the fastcgi bufferevent while we're looping. We don't support fastcgi multiplexing, so once we get an END_REQUEST there's nothing more to do. Prodded into looking here after a bug report from Allen Sobot, thanks!

The code in fcgi_req to send the custom params set in the config file was placed inside the conditional for `tls_peer_cert_provided`, so the custom parameters would not be sent if a client certificate is not provided.

matches found with % grep -R '=[ ]*{' . | fgrep -v const

FastCGI is designed to multiplex requests over a single connection, so ideally the server can open only one connection per worker to the FastCGI application and that's that. Doing this kind of multiplexing makes the code harder to follow and easier to break/leak etc on the gmid side however. OpenBSD' httpd seems to open one connection per client, so why can't we too? One connection per request is still way better (lighter) than using CGI, and we can avoid all the pitfalls of the multiplexing (keeping track of "live ids", properly shut down etc...)

We ended up copying too much data from the fastcgi process.

This is a big change in how gmid handles I/O. Initially we used a hand-written loop over poll(2), that then was evolved into something powered by libevent basic API. This meant that there were a lot of small "asynchronous" function that did one step, eventually scheduling the re-execution, that called each others in a chain. The new implementation revolves completely around libevent' bufferevents. It's more clear, as everything is implemented around the client_read and client_write functions. There is still space for improvements, like adding timeouts for one, but it's solid enough to be committed as is and then further improved.

add the reported request id if there's a mismatch and both the gai error and the errno value if getnameinfo fails.