Anarthal (Rubén Pérez)
Anarthal (Rubén Pérez)
Following https://github.com/boostorg/system/issues/115, review that we're not affected, too.
* The retry strategy is currently very simple (linear time between retries) and not configurable. When there are a lot of connections, this can saturate the network. A configurable retry...
This is a rare case of a race condition that can happen during the execution of any connection function. If the cancellation signal is emitted while the current async op...
Per-operation cancellation for `async_run` and `async_get_connection` doesn't do the right thing. It should: * `async_run`: same thing as `connection_pool::cancel`. This is partial cancellation. * `async_get_connection`: mark the operation as cancelled,...
All our CI systems have `BOOST_ASIO_HAS_LOCAL_SOCKETS` defined, but some legacy systems may not support it. Include a build that makes `BOOST_ASIO_HAS_LOCAL_SOCKETS` false.
The new `any_connection` doesn't allow custom `Stream` types - it always uses `detail::variant_stream`. I'm not sure whether this is really required, since this stream is designed to cover almost all...
TSAN tests were disabled due to a reported race condition in Asio for systems using kqueue (like OSX). Same stacktrace as https://github.com/chriskohlhoff/asio/issues/641 (except we're not mixing sync and async).
`connection_pool` doesn't allow setting metadata mode, so there's no way to retrieve metadata strings from a pooled connection (the static interface still works, since it bypasses metadata mode). Having a...
There is no reason why handshake_params couldn't have all member functions constexpr.