Strange 'Connection closed while waiting for data.' Zinc error

> On 25 Dec 2017, at 05:42, Andrew Glynn <[hidden email]> wrote:
>
> The only time I've seen that behaviour is with the combination of a slow server and a proxy such as CloudFlare where the sysadmin has set tcp_wait to some ridiculously small number.  Have you tried setting the connection reuse timeout to 0, or setting 'beOneShot'?
>
> Zinc doesn't support the additions in HTTP/2, but I can't offhand think of any reason that should cause that specific behaviour. I've had sufficient problems with HTTP/2 over slow connections (I live where there are no land lines of any kind) that I replaced the HTTP/2 code in Nginx with code from pureftpd in order to access my server at home when I'm out.
>
> Andrew
>
> On 2017-12-24, 3:30 PM, "Pharo-users on behalf of Yuriy Tymchuk" <[hidden email] on behalf of [hidden email]> wrote:
>
>    Hi,
>
>    I’m trying to hack something with Pharo 6.1 and Phillips Hue[1] (they have a simple REST API). When I try to do a GET request with Zinc like this:
>
>    ZnClient new
>     url: 'http://192.168.0.115:80/api/1028d66426293e821ecfd9ef1a0731df/lights/';
>     get
>
>    I get a “ConnectionClosed: Connection closed while waiting for data” exception. The service provides a JSON response that I can get without a problem by using a web browser or curl. Also if I try to access fewer data (e.g. adding /1 to the url, so it gets only one resource) I don’t get the exception.
>
>    What is even more strange, when I resume the exception I get the desired response i.e. this works:
>
>    [ ZnClient new
>     url: 'http://192.168.0.115:80/api/1028d66426293e821ecfd9ef1a0731df/lights/';
>     get ]
>
>     on: ConnectionClosed
>     do: [ :ex |
>     ex resumeUnchecked: ex defaultResumeValue ]
>
>    Of course I cannot share the server that is on my local network for resting, but maybe anyone knows it there are any known issues with Zinc, or if there are ways for me to debug the issue further.
>
>    Cheers.
>    Uko
>
>    [1]: https://www.developers.meethue.com
>
>
>
>

> Cheers,
> Uko
>
>> On 25 Dec 2017, at 05:42, Andrew Glynn <[hidden email]> wrote:
>>
>> The only time I've seen that behaviour is with the combination of a slow server and a proxy such as CloudFlare where the sysadmin has set tcp_wait to some ridiculously small number.  Have you tried setting the connection reuse timeout to 0, or setting 'beOneShot'?
>>
>> Zinc doesn't support the additions in HTTP/2, but I can't offhand think of any reason that should cause that specific behaviour. I've had sufficient problems with HTTP/2 over slow connections (I live where there are no land lines of any kind) that I replaced the HTTP/2 code in Nginx with code from pureftpd in order to access my server at home when I'm out.
>>
>> Andrew
>>
>> On 2017-12-24, 3:30 PM, "Pharo-users on behalf of Yuriy Tymchuk" <[hidden email] on behalf of [hidden email]> wrote:
>>
>>   Hi,
>>
>>   I’m trying to hack something with Pharo 6.1 and Phillips Hue[1] (they have a simple REST API). When I try to do a GET request with Zinc like this:
>>
>>   ZnClient new
>>   url: 'http://192.168.0.115:80/api/1028d66426293e821ecfd9ef1a0731df/lights/';
>>   get
>>
>>   I get a “ConnectionClosed: Connection closed while waiting for data” exception. The service provides a JSON response that I can get without a problem by using a web browser or curl. Also if I try to access fewer data (e.g. adding /1 to the url, so it gets only one resource) I don’t get the exception.
>>
>>   What is even more strange, when I resume the exception I get the desired response i.e. this works:
>>
>>   [ ZnClient new
>>   url: 'http://192.168.0.115:80/api/1028d66426293e821ecfd9ef1a0731df/lights/';
>>   get ]
>>
>>   on: ConnectionClosed
>>   do: [ :ex |
>>   ex resumeUnchecked: ex defaultResumeValue ]
>>
>>   Of course I cannot share the server that is on my local network for resting, but maybe anyone knows it there are any known issues with Zinc, or if there are ways for me to debug the issue further.
>>
>>   Cheers.
>>   Uko
>>
>>   [1]: https://www.developers.meethue.com
>>
>>
>>
>>
>
>

> On 25 Dec 2017, at 20:03, Sven Van Caekenberghe <[hidden email]> wrote:
>
>
>
>> On 25 Dec 2017, at 18:03, Yuriy Tymchuk <[hidden email]> wrote:
>>
>> Hi,
>>
>> setting the reuse timeout to 0 or making a one shot client does not help. But while googling I found the exact issue described on the Pharo issue tracker: https://pharo.fogbugz.com/f/cases/18588/
>>
>> The proposed workaround also worked for me:
>> ZnNetworkingUtils default socketStreamClass: SocketStream.
>
> You are on Windows I guess ?
>
>> Cheers,
>> Uko
>>
>>> On 25 Dec 2017, at 05:42, Andrew Glynn <[hidden email]> wrote:
>>>
>>> The only time I've seen that behaviour is with the combination of a slow server and a proxy such as CloudFlare where the sysadmin has set tcp_wait to some ridiculously small number.  Have you tried setting the connection reuse timeout to 0, or setting 'beOneShot'?
>>>
>>> Zinc doesn't support the additions in HTTP/2, but I can't offhand think of any reason that should cause that specific behaviour. I've had sufficient problems with HTTP/2 over slow connections (I live where there are no land lines of any kind) that I replaced the HTTP/2 code in Nginx with code from pureftpd in order to access my server at home when I'm out.
>>>
>>> Andrew
>>>
>>> On 2017-12-24, 3:30 PM, "Pharo-users on behalf of Yuriy Tymchuk" <[hidden email] on behalf of [hidden email]> wrote:
>>>
>>>  Hi,
>>>
>>>  I’m trying to hack something with Pharo 6.1 and Phillips Hue[1] (they have a simple REST API). When I try to do a GET request with Zinc like this:
>>>
>>>  ZnClient new
>>>   url: 'http://192.168.0.115:80/api/1028d66426293e821ecfd9ef1a0731df/lights/';
>>>   get
>>>
>>>  I get a “ConnectionClosed: Connection closed while waiting for data” exception. The service provides a JSON response that I can get without a problem by using a web browser or curl. Also if I try to access fewer data (e.g. adding /1 to the url, so it gets only one resource) I don’t get the exception.
>>>
>>>  What is even more strange, when I resume the exception I get the desired response i.e. this works:
>>>
>>>  [ ZnClient new
>>>   url: 'http://192.168.0.115:80/api/1028d66426293e821ecfd9ef1a0731df/lights/';
>>>   get ]
>>>
>>>   on: ConnectionClosed
>>>   do: [ :ex |
>>>   ex resumeUnchecked: ex defaultResumeValue ]
>>>
>>>  Of course I cannot share the server that is on my local network for resting, but maybe anyone knows it there are any known issues with Zinc, or if there are ways for me to debug the issue further.
>>>
>>>  Cheers.
>>>  Uko
>>>
>>>  [1]: https://www.developers.meethue.com
>>>
>>>
>>>
>>>
>>
>>
>
>

> Cheers.
> Uko
>
>> On 25 Dec 2017, at 20:03, Sven Van Caekenberghe <[hidden email]> wrote:
>>
>>
>>
>>> On 25 Dec 2017, at 18:03, Yuriy Tymchuk <[hidden email]> wrote:
>>>
>>> Hi,
>>>
>>> setting the reuse timeout to 0 or making a one shot client does not help. But while googling I found the exact issue described on the Pharo issue tracker: https://pharo.fogbugz.com/f/cases/18588/
>>>
>>> The proposed workaround also worked for me:
>>> ZnNetworkingUtils default socketStreamClass: SocketStream.
>>
>> You are on Windows I guess ?
>>
>>> Cheers,
>>> Uko
>>>
>>>> On 25 Dec 2017, at 05:42, Andrew Glynn <[hidden email]> wrote:
>>>>
>>>> The only time I've seen that behaviour is with the combination of a slow server and a proxy such as CloudFlare where the sysadmin has set tcp_wait to some ridiculously small number.  Have you tried setting the connection reuse timeout to 0, or setting 'beOneShot'?
>>>>
>>>> Zinc doesn't support the additions in HTTP/2, but I can't offhand think of any reason that should cause that specific behaviour. I've had sufficient problems with HTTP/2 over slow connections (I live where there are no land lines of any kind) that I replaced the HTTP/2 code in Nginx with code from pureftpd in order to access my server at home when I'm out.
>>>>
>>>> Andrew
>>>>
>>>> On 2017-12-24, 3:30 PM, "Pharo-users on behalf of Yuriy Tymchuk" <[hidden email] on behalf of [hidden email]> wrote:
>>>>
>>>> Hi,
>>>>
>>>> I’m trying to hack something with Pharo 6.1 and Phillips Hue[1] (they have a simple REST API). When I try to do a GET request with Zinc like this:
>>>>
>>>> ZnClient new
>>>> url: 'http://192.168.0.115:80/api/1028d66426293e821ecfd9ef1a0731df/lights/';
>>>> get
>>>>
>>>> I get a “ConnectionClosed: Connection closed while waiting for data” exception. The service provides a JSON response that I can get without a problem by using a web browser or curl. Also if I try to access fewer data (e.g. adding /1 to the url, so it gets only one resource) I don’t get the exception.
>>>>
>>>> What is even more strange, when I resume the exception I get the desired response i.e. this works:
>>>>
>>>> [ ZnClient new
>>>> url: 'http://192.168.0.115:80/api/1028d66426293e821ecfd9ef1a0731df/lights/';
>>>> get ]
>>>>
>>>> on: ConnectionClosed
>>>> do: [ :ex |
>>>> ex resumeUnchecked: ex defaultResumeValue ]
>>>>
>>>> Of course I cannot share the server that is on my local network for resting, but maybe anyone knows it there are any known issues with Zinc, or if there are ways for me to debug the issue further.
>>>>
>>>> Cheers.
>>>> Uko
>>>>
>>>> [1]: https://www.developers.meethue.com
>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>
>