Re: Threaded fastcgi-server - truncated output

Vladimir,

There is a buffering going on in the FastCGI library between the application and 
the server.  It sounds like something is either interacting with that, or with 
something else.  I don't think any of our perl code (using CGI::Fast) shows a 
sleep going on, but it's also not threaded.

Any lower-level developers out there who can comment on this?  Sounds like 
either a code patch, documentation patch, or both might be due.

David.

-----

On Sat, 13 Oct 2007, Vladimir V. Pavluk wrote:

> Hello!
>
> Playing with my code making it similar to the example I figured out that the
> problem appears when I'm setting the O_NONBLOCK flag on the socket which I
> listen to.
> Is it a bug, or it is a documented behaviour?
>
> Sincerely, Vladimir V.Pavluk.
>
>> Hello!
>>
>> I've wrote the threaded fastcgi application, which is generally works fine.
>> However, I've found a problem: when the size of output becomes more than
>> some threshold (in my case - about 8.7 kb), the rest output is truncated.
>> It seemed to me that the stream does not have time to be flushed before the
>> FCGX_Finish_r is called, because when I insert sleep(1) line, the problem
>> is gone.
>> I've found an example of threaded fastcgi server in dev kit distribution,
>> and found the 'sleep(2)' line just after output.
>> So, I have following questions:
>> Is this delay really needed for allowing streams to be flushed?
>> How can we do it without sleep delays (if it is possible at all) - I don't
>> really want to insert sleeps after output as it looks as a dirty hack :(
>> ?..
>>
>> Sincerely, Vladimir V. Pavluk.
>> ___________________________________
>> fastcgi-developers mailing list
>> http://fastcgi.com/fastcgi-developers/
>
>
> ___________________________________
> fastcgi-developers mailing list
> http://fastcgi.com/fastcgi-developers/
>
___________________________________
fastcgi-developers mailing list
http://fastcgi.com/fastcgi-developers/