REQ not sending ETIMEDOUT

[thelinuxlich]

According to @gdamore on nanomsg gitter channel:

REQ will return ETIMEDOUT on recv() if the timeout expires.

So far I've tried every configuration possible and req.on("timeout", function() { ... }) never fires.

[gdamore]

Make sure the NN_RCVTIMEO option is set though!

On Jun 1, 2015, at 2:17 PM, Alisson Cavalcante Agiani notifications@github.com wrote:

According to @gdamore https://github.com/gdamore on nanomsg gitter channel:

REQ will return ETIMEDOUT on recv() if the timeout expires.
So far I've tried every configuration possible and req.on("timeout", function() { ... }) never fires.

—
Reply to this email directly or view it on GitHub #93.

[thelinuxlich]

Oh yeah, I've tried setting it to 1, nothing happened

[nickdesaulniers]

@thelinuxlich do you have a small test case we could add to our test suite to work against?

[gdamore]

Btw, ETIMEOUT does not apply to RAW sockets -- only to cooked ones.

[gdamore]

Huh. In my testing I'm getting EAGAIN instead of ETIMEDOUT. That's ... not what it should be.

[nickdesaulniers]

@gdamore , note we statically link an older version of Nanomsg. If that was recently fixed, we can update our dependency.

[gdamore]

Huh. In libnanomsg ETIMEDOUT is intentionally turned into EAGAIN:
rc = nn_efd_wait (&self->rcvfd, timeout);
if (nn_slow (rc == -ETIMEDOUT))
return -EAGAIN;

This feels like a mistake to me, but I need to make further research. Nonetheless, EAGAIN or ETIMEDOUT -- you do get a timeout. Blocking mode only of course. (nonblocking sockets cannot distinguish a "timeout" as such.)

[thelinuxlich]

hmmm req.send is non-blocking in node-nanomsg? If it is we are lost concerning timeouts? :(

[thelinuxlich]

@nickdesaulniers I'm trying to get the timeout event with this simple server implementation: https://gist.github.com/thelinuxlich/1e55ef121874d127ceca

[nickdesaulniers]

Well, you're missing a rep socket. Req/rep are meant to be used as a pair. I should add an example for this...

[nickdesaulniers]

for now, see this test

[thelinuxlich]

no, I'm testing exactly the absence of a consumer listening to see a timeout event

[nickdesaulniers]

indeed, we do not emit a timeout event anywhere. Note the ability to specify socket options were added recently, and aren't fully tested other than check for their existence. See #52. What function in nanomsg returns ETIMEOUT or EAGAIN? That would help me figure out how to wire it up.

[nickdesaulniers]

Stack looks like nn_send -> nn_sendmsg -> nn_sock_send -> EAGAIN

[thelinuxlich]

@gdamore said REQ is sending EAGAIN

[gdamore]

nn_recv gets the EAGAIN.

Sent from my iPhone

On Jun 1, 2015, at 3:17 PM, Nick Desaulniers notifications@github.com wrote:

Stack looks like nn_send -> nn_sendmsg -> nn_sock_send -> EAGAIN

—
Reply to this email directly or view it on GitHub.