Conversation
Minimize the number of layers in Dockerfiles. Add buildx caches for Ruby, Crystal, Python, and Go. Build Gleam framework images in a multi-stage build. Specify the locale `C.UTF-8` in Elixir Dockerfile - Elixir issued a warning that the locale must be specified manually.
Update hyperlane
bump version
Update dependency ch.qos.logback:logback-classic to v1.6.2
Update dependency @nifrajs/node to ~2.12.0
Update spider-gazelle
Update dependency django to >=6,<6.2
Update dart Docker tag to v3.13
Update dependency org.restheart:restheart to v9.7.2
The benchmark suite uses oha with --disable-keepalive (see .tasks/config.rake), so every request opens a new TCP connection. The previous 21.3 handler still had three constructs that hurt this scenario: - .set_nodelay(Some(false)) enabled Nagle's algorithm, adding ~40ms to the small responses used in this benchmark. - set_header(CONNECTION, KEEP_ALIVE) was useless under disable-keepalive and added a buffer write per request. - A method re-check inside each handler (the route is already matched). - Manual stream.set_closed(true) + Status::Reject on try_send failure. - RequestConfig::low_security() added request-side checks with no benefit on this benchmark. This trims the handler to the minimum needed to satisfy .spec/route_spec.rb (status 200, correct body). Net -31 lines. Note: this drops the manual keep-alive while-loop that existed in the 20.2 implementation; 21.3 already removed that loop, so disable-keepalive only loses the headers / method-check / Nagle work here. Behavior is unchanged on the spec contract: - GET / -> 200, empty body - GET /user/0 -> 200, body '0' - POST /user -> 200, empty body
[Rust] Optimize hyperlane handler for disable-keepalive benchmark
Update dependency ch.qos.logback:logback-classic to v1.6.3
…lications-helidon-se-4.x Update dependency io.helidon.applications:helidon-se to v4.5.3
Update dependency dev.zio:zio-http to v3.11.4
|
@niden Seems that https://github.com/phalcon/cphalcon/releases/download/v5.18.2/phalcon-pecl.tgz is in fact the source code of 5.18.1 |
It is the same code - there was literally no change between .1 and .2. .2 had an change in the build artifacts i.e. not to send the whole repo to the installer but only the files required. I forgot to run the bump script to make that .2, and this is why it still reports .1 This will be fixed today with the new release coming up. |
|
@olliNiinivaara @bung87 @ire4ever1190 could you take a look at while compiling it |
|
@enghitalo could you check veb, I have a compilation issue PS : Same issue for vanilla epoll |
Fix guildenstern, use `lvlNone` instead of `NONE`
|
@josephmancuso With could you take a look ? |
|
@fenric For Could you take a look ? |
Failed frameworks