UPDATE: The bug has been [fixed in
devel](https://github.com/nim-lang/Nim/pull/15012) thanks to @dom96. Turned out
to be a mistake with not handling EAGAIN properly. I am much relieved.
Same assertion error on Linux.
...
RCVD #269: 000269 This is message #269 of ∞. Please stay tuned for more.
000269 This is me...
RCVD #270: 000270 This is message #270 of ∞. Please stay tuned for more.
000270 This is me...
RCVD #715: 000715 This is message #715 of ∞.
In case you haven't got feedback from Linux users, this is running the gh test
case under Linux (Debian bullseye, kernel 5.7:
[...]
SENT #18336
SENT #18337
SENT #18338
SENT #18339
SENT #18340
RCVD #647: 000647 This is message #647 of ∞. Please stay tuned for
Filed
[https://github.com/nim-lang/Nim/issues/15003](https://github.com/nim-lang/Nim/issues/15003)
Would appreciate it if anyone could run the test case (there's a newer version
attached to the Github issue) on non-Mac systems and see if it reproduces. The
bug may well be specific to kqueues in
Here's the plain AsyncSocket version. The message counts are different but the
failure is the same.
import asyncdispatch, asyncnet, strformat, sugar
const FrameSize = 999
proc runServer() {.async.} =
# Server side:
var server = newAsyncSocket()
OK, here's a reproducible test case (still using news). It opens a loopback
socket; the client side sends out messages as fast as possible, and the server
reads the messages slowly, to create some backpressure. The client watches the
order that the send futures complete, to make sure they're con
I wasn't sure which WebSocket module to use ... news seems like the newest,
since it's based on treeform/ws which is based on niv/websocket ¯_(ツ)_/¯ All
three libraries are pretty immature,
I don't see how this can be news's fault, since if you look at [the send()
function](https://github.com/T
Wow, we now have 3 different websocket implementations. Nice :)
I must say I am suspicious of `news` here, if I were you I would do the
following:
* Check whether the same behaviour happens with Chronos (since `news`
apparently supports it as well)
* Reproduce this problem with an `AsyncSoc
**tl;dr:** I'm issuing lots of small writes to an AsyncSocket, and some of the
writes are being dropped — they don't show up on the other end of the TCP
connection, and their corresponding `send` Futures never complete.
I'm using the `news` package to send lots of small WebSocket messages (on th
