Use HttpRecord as response body so requests can be tracked all the way
to response body completion.
This allows Request properties to be accessed while the response body is
streaming.
Graceful shutdown now awaits a future instead of async spinning waiting
for requests to finish.
On the minimal benchmark this refactor improves performance an
additional 2% over pooling alone for a net 3% increase over the previous
deno main branch.
Builds upon https://github.com/denoland/deno/pull/20809 and
https://github.com/denoland/deno/pull/20770.
---------
Co-authored-by: Matt Mastracci <matthew@mastracci.com>
Reuse existing existing allocations for HttpRecord and response
HeaderMap where possible.
At request end used allocations are returned to the pool and the pool
and the pool sized to 1/8th the current number of inflight requests.
For http1 hyper will reuse the response HeaderMap for the following
request on the connection.
Builds upon https://github.com/denoland/deno/pull/20770
---------
Co-authored-by: Matt Mastracci <matthew@mastracci.com>
Makes the JavaScript Request use a v8:External opaque pointer to
directly refer to the Rust HttpRecord.
The HttpRecord is now reference counted. To avoid leaks the strong count
is checked at request completion.
Performance seems unchanged on the minimal benchmark. 118614 req/s this
branch vs 118564 req/s on main, but variance between runs on my laptop
is pretty high.
---------
Co-authored-by: Matt Mastracci <matthew@mastracci.com>
This PR uses the new `cancel` method of `TransformStream` to properly
clean up the internal `TextDecoder` used in `TextDecoderStream` if the
stream is cancelled.
Fixes #13142
Co-authored-by: Bartek Iwańczuk <biwanczuk@gmail.com>
We only want one zlib dependency.
Zlib dependencies are reorganized so they use a hidden
`__vendored_zlib_ng` flag in cli that enables zlib-ng for both libz-sys
(used by ext/node) and flate2 (used by deno_web).
This also updates deno_graph, which has the JSR change to use "exports".
It's not yet useful atm, so I've made this PR a fix about the deno doc
--lint error message improvements. I'll do a follow-up PR that adds
exports to the deno.json
This PR reduces the time to run `queue persistence with inflight
messages` test from 60 seconds down to about 6 seconds.
The test simulates a crash to ensure that inflight messages are cleaned
up and re-queued when the new process starts. Since messages are
considered dead after 5 seconds of being queued, reopening the db within
5 seconds does not re-queue the messages on startup (they do get
re-queued after 60 seconds, which is the cleanup frequency). By waiting
for 5 seconds before reopening the db, the test ensures that the cleanup
happens quickly when the db is opened.
We can move all promise ID knowledge to deno_core, allowing us to better
experiment with promise implementation in deno_core.
`{un,}refOpPromise(promise)` is equivalent to
`{un,}refOp(promise[promiseIdSymbol])`
Remove tokio-rustls as a direct dependency of Deno and refactor
test_server to reduce code duplication.
All tcp and tls listener paths go through the same streams now, with the
exception of the simpler Hyper http-only handlers (those can be done in
a later follow-up).
Minor bugs fixed:
- gRPC server should only serve h2
- WebSocket over http/2 had a port overlap
- Restored missing eye-catchers for some servers (still missing on Hyper
ones)
