lwip_sock: re-issue receive events if there are still received messages after *_recv() or *_accept() was called.#14060
Conversation
... after `*_recv()` or `*_accept()` was called. This prevents a client from needing to call `*_recv()` or `*_accept()` in a loop.
|
CI is broken right now, all tests run into timeouts leading to a failed job after ~2h |
|
Ping? We should merge this soonish so the APIs with GNRC and lwIP don't get too much out of sync. |
|
I'd be happy to test but I'm not sure how to properly build e.g. |
|
Mh, I know @kb2ma had gcoap already running. Let me have a look if I can find it. |
MichelRottleuthner
left a comment
There was a problem hiding this comment.
Even with #13641 it doesn't seem possible to use the same procedure for testing as I described in #14034 because CoAp simply doesn't work for me with that. But that is probably related to #13641 itself.
Using tests/lwip I can confirm that using bare UDP and TCP with sock_async still works and this PR gives less dropped packets on fast incoming bursts. ACK.
Contribution description
This prevents a client from needing to call
*_recv()or*_accept()in a loop.Testing procedure
See #14034, but with
USEPKG = lwipinstead of using GNRC.Issues/PRs references
Fixes #14034 when used with lwIP.
See also #14059 for the GNRC variant of this fix.