It was decided that the performance benefits that isolates offer (faster spin-up
times for worker processes, faster inter-worker communication, possibly a lower
memory footprint) are not actual bottlenecks for most people and do not outweigh
the potential stability issues and intrusive changes to the code base that
first-class support for isolates requires.
Hence, this commit backs out all isolates-related changes.
Good bye, isolates. We hardly knew ye.
- Save StringPtr if the header hasn't been completely received yet after one
packet.
- Add one to num_fields and num_values. They were actually one less than the
number of fields and values.
- Remove always_inline makes debugging difficult, and has negligible
performance benefits.
There is a difference between errors which happen to a socket - like
receiving EPIPE - an exceptional situation but ultimately okay and the
situation where code throws in a callback - which is not okay.
Fixes test/simple/test-http-exceptions.js
TODO: explain this in docs.
This allows for web servers to be "hijacked" and used as Web Socket servers
(or other). You simply listen for requests as normal, but check if
req.upgrade === true
If so, this will be the last request of the connection. It's your job now to
hijack req.connection and start reading from it. req.upgradeHead is a buffer
containing the first part of the new protocol communication (in the case it
arrived on the same packet).
This needs tests and documentation. API subject to change.
However it's not working very well: Hitting a 'hello world' server with many
requests (ab -t 60 -c 10) will cause it to crash with the following error.
Obtained 3 stack frames.
./node(_Z11print_tracev+0x1c) [0x80d1b3c]
./node(_ZN4node6Parser7ExecuteERKN2v89ArgumentsE+0x69) [0x80d3759]
./node [0x811f44b]
TypeError: Already parsing a buffer
at Socket.<anonymous> (/home/ryan/projects/node/lib/http2.js:393:20)
at IOWatcher.callback (/home/ryan/projects/node/lib/net.js:81:12)
at node.js:985:9
at node.js:989:1