#grpc

does anyone has any idea about the Any type as in protobuf 3?

Removal of extensions, which are instead replaced by a new standard type called Any.

hallo

hello

heya

Yesterday I left the office happily thinking I'd fixed the tsan issues

today not so much

but I got back my backpack which I left on the subway, so I've got that going for me, which is nice

:D

good mornigs

trying to wrap my head around async calls in grpc/c++

the whole idea behind CompletionQueue and CompletionQueueTag interface still eludes me

hi kolmodin_

hi Kanalia :)

I seem to have a pretty poor connection through my ISP at home. ( that's kolmodin_)

it happens even to the best of them

I got disconnected today by freenode too, even though I got a good connection

yay, this should help me in implementing async client in grpc: https://github.com/grpc/grpc/issues/899

JohanSJA: Any is a message that is basically just a string type and a byte[] content.

hi ejona

Howdy, Kanalia.

|Pixel|: are you around?

ejona: how to cleanly disconnect from grpc-java server ?

I always get a "java.io.IOException: An existing connection was forcibly closed by the remote host" throw when my client disconnects from the server

Hmm...

On Java side, ChannelImpl.shutdown() would do it.

I don't know about C side.

ok

Want to create an issue and include the full exception you get, so I can find out if it is a Java problem on the C could do a cleaner termination?

nah, not ATM, let's wait for |Pixel|

+ I have to register a github account, finally :P

has nyone gotten grpc csharp to run on Visual Studio

solfried: I haven't tried

docs say that the csharp version is currently built on Mono, but I am trying to run on Windows

I have gotten a successfull compile and can run the server

it crashes when receiving messages

that guy didn't have much patience it seems

cool, now I got working async client (at least I think so)

:( I would have told him that I think C# is still in the works. https://github.com/grpc/grpc/tree/master/src/cs... says it only works on Mono right now, and even then "This gRPC C# implementation is work-in-progress and is not expected to work yet."

yes, pre-alpha stage after all

Ah, he's now sent an email to grpc-io, so I can still inform him.

yay

still, typical case of impatient web-irc user :P

Kanalia: what's up ?

|Pixel|: hi Nicolas

heya

how to cleanly/properly disconnect from the server in grpc/grpc++ ?

well, you shutdown everything

been trying to find some method of closing the channel/socket/connection before calling grpc_shutdown() but I wasn't successful

just that?

ok

it's refcounted

so if you shutdown all of the completion queues

and shutdown grpc itself

yes I shutdown everything

all of the refcounts should go to 0

and things should clear

still, in such a situation, grpc-java server throws an exception like this: "java.io.IOException: An existing connection was forcibly closed by the remote host"

hmm

we were wondering with ejona if that's a problem on the java side or am I doing something wrong on the client side

it may be the first case then

I don't recall the protocol having an actual shutdown handshake ?

maybe it doesn't, too bad ejona parted, will have to talk to him about this later

because I wonder how it works when java talks with java

but that warning doesn't make me cringe, no

nah it's nothing really bad

what do you mean ?

just inconvenient in the logs :P you don't know if really an IOException happened at one point or if it's just the client disconnecting

oh right

I meant that I wonder if java client talking to java server, then disconnecting, produces the same IO exception

if not, then it's obviously an implementation fail, the server expects some notice but doesn't get it from a c/cpp client

yeah, it's just speculation now, ejona will know more about this on that side

still I managed to get the hang of async connection on grpc with the QPS tests

I haven't spelunked the code enough yet

but I don't recall a goodbye message

or any mechanism towards that for that matter

I didn't understand the whole notions with tags and the complettion queue methods, but after a few try&error sessions it all made sense

ok, good to know

ah, one more question, how to disable all this verbose debugging output that's thrown at std::err by grpc ?

you can override the logger... :)

ah so those are gpr_log calls?

it's a bit, well

non intuitive

yes

ok, that's enough for me to handle it

void gpr_set_log_function(gpr_log_func func);

yay, thanks

what's the diff between completionqueue Next(...) and AsyncNext(...) methods?

that AsyncNext works on a deadline and is non-blocking, while next is blocking?

that should be it, yes

ok

oh and finally recompiled with your latest windows fixes, gprc doesn't crash now on finish

*grpc

cool :)