Re: CHUNKSIZE in connector.c

[Chris Townsend <Christopher.Townsend@xxxxxxxxxxxxx>]

Hi Andreas,

Sure, will do.

Chris

On 7/29/19 1:18 PM, Andreas Schneider wrote:
> On Friday, July 19, 2019 2:58:11 PM CEST Chris Townsend wrote:
>> Hey Andreas,
>>
>> Ok, I agree a proper fix is the best route, particularly for the libssh
>> project.
> Chris,
>
> could you open a bug report?
>
>> That said, we have a number of users that are being impacted by this so
>> I'll apply the workaround to our project's libssh and then, as time
>> permits, I can work on trying to fix this properly.
>>
>> Thanks for your input!
>>
>> Chris
>>
>> On 7/19/19 8:42 AM, Andreas Schneider wrote:
>>> On Friday, July 19, 2019 1:53:30 PM CEST Chris Townsend wrote:
>>>> Hi Andreas,
>>> Hi Chris,
>>>
>>>> Well, I did try 32768 bytes for CHUNKSIZE and although the bug was much
>>>> less likely to happen, I still did see it occur.  Maybe there was some
>>>> overhead in the frame or some such, so I thought 64K should be enough:)
>>>>
>>>> Also, I do agree that it just hides the problem, so I did try to fix it
>>>> via a loop to read again from the socket, but the problem I ran into was
>>>> that it's not clear if the next read is for this particular event or for
>>>> another event that is not related.  I then started to read data for
>>>> other events and that was getting outputted to the screen via the
>>>> socket->stdout connector.
>>> I think we need a test for this to reproduce and then fix it correctly. If
>>> it fails in this case I'm sure there are others.
>>>
>>> So if you could look into that, that would be great, else file a bug
>>> report.
>>>
>>>
>>> Thanks,
>>>
>>> 	Andreas
>