If you have multiple vhosts with client contexts enabled, under
OpenSSL each one brings in the system cert bundle.
On libwebsockets.org, there are many vhosts and the waste adds up
to about 9MB of heap.
This patch makes a sha256 from the client context configuration, and
if a suitable client context already exists on another vhost, bumps
a refcount and reuses the client context.
In the case client contexts are configured differently, a new one
is created (and is available for reuse as well).
info.protocols works okay, but it has an annoying problem... you have to know
the type for each protocol's pss at the top level of the code, so you can set
the struct lws_protocols user_data size for it.
Lws already rewrites the protocol tables for a vhost in the case of runtime
protocol plugins... this adapts that already-existing code slightly to give
a new optional way to declare the protocol array.
Everything works as before by default, but now info.protocols may be NULL and
info.pprotocols defined instead (if that's also NULL, as it will be if you
just ignore it after memsetting to 0, then it continues to fall back to the
dummy protocol handler as before).
info.pprotocols is a NULL-termined array of pointers to lws_protocol
structs. This can be composed at the top level of your code without knowing
anything except the name of the externally-defined lws_protocol struct(s).
The minimal example http-server-dynamic is changed to use the new scheme as
an example.
With SMP as soon as we add the new sockfd to the fds table, in the
case we load-balanced the fd on to a different pt, service on it
becomes live immediately and concurrently. This can lead to the
unexpected situation that while we think we're still initing the
new wsi from our thread, it can have lived out its whole life
concurrently from another service thread.
Add a volatile flag to inform the owning pt that if it wants to
service the wsi during this init window, it must wait and retry
next time around the event loop.
On h1, cgi stdout close doesn't prompt the http close, instead it
times out. Fix that so we also close on h1, and make the close
action itself on http timeout less drastic.
As it was, GnuTLS actually marks the close as a fatal TLS error.
