Ben,

> On 18 Feb 2018, at 15:29, Ben Coman <[email protected]> wrote:
> 
> The websockets guide here....
> https://ci.inria.fr/pharo-contribution/job/EnterprisePharoBook/lastSuccessfulBuild/artifact/book-result/WebSockets/WebSockets.html
> 
> says "Reading and sending are completely separate and independent" and the 
> class comment says its "full-duplex".  From these I presume sending and 
> receiving are okay to occur in separate threads, but this is not explicitly 
> stated, so can someone confirm this?

The reading and the writing over a web socket connection are independent, so 
yes that would be OK. Of course, they would both share the same network stream 
(just use each half independently) - that normally works fine.

> So for the given example....
> | webSocket |
> webSocket := ZnWebSocket to: 'ws://echo.websocket.org'.
> [ webSocket
>    sendMessage: 'Pharo Smalltalk using Zinc WebSockets !';
>    readMessage ] ensure: [ webSocket close ].
> 
> re-implementing as follows seems to work...
> webSocket := ZnWebSocket to: 'ws://echo.websocket.org'.
> [ webSocket runWith: [ :msg | Transcript crShow: msg ] ] forkAt: 35.
> webSocket sendMessage: 'Pharo Smalltalk using Zinc WebSockets !'.
> webSocket sendMessage: 'Another multi-thread echo'.
> webSocket close
> 
> but that could just be by luck, so I ask.

No, that should work.

> btw, if I don't send the #close in the last line, after ~30 seconds I get an 
> error
>     ConnectionClosed: Cannot write data
> which seems reasonable that the other end timed out and closed the connection,
> 
> but when I send the #close, after ~30 seconds I get an error...
>    PrimitiveFailed: primitive #primSocketReceiveDataAvailable: in Socket 
> failed
> which is unfriendly.  What is the best way to neatly stop trying to receive 
> data when we explicitly close the websocket?

Of course that fails, you fork a reading loop containing #runWith: which has to 
end/stop somehow, the lazy way is by allowing an error to bubble up.

> cheers -ben
> 
> 
> P.S.  I wonder if when a ZnWebSocket is closed, it might be worthwhile
> to do  "role:=#closed"  to provide some visibility of its state in its 
> GTInspector [Raw] tab..

Maybe, but there is #isConnected and the Socket inspector 'Socket Info' shows 
lots of details, you can rely on that.

Sven


Reply via email to