2009/2/10 Graham Dumpleton <[email protected]>:
> 2009/2/10 Graham Dumpleton <[email protected]>:
>> 2009/2/10 Graham Dumpleton <[email protected]>:
>>> 2009/2/10 Porcari Giovanni <[email protected]>:
>>>>
>>>> Il giorno 10/feb/09, alle ore 01:05, Graham Dumpleton ha scritto:
>>>>
>>>>> Giovanni, if you are still reading, would it be possible for you to
>>>>> try with Apache 2.2.9 and Python 2.5.2 that ship with MacOS X
>>>>> (Leopard) instead of your separate versions.
>>>>>
>>>>> Ie., for mod_wsgi use:
>>>>>
>>>>> ./configure --with-apxs=/usr/sbin/apxs --with-python=/usr/bin/python
>>>>>
>>>>> if PATH is picking up other versions first.
>>>>>
>>>>> At the moment the thing that links both of you is that you are using
>>>>> Apache 2.2.11. My quick test on older version supplied with MacOS X
>>>>> works fine.
>>>>>
>>>>> I will have to try and install Apache 2.2.11 and see if I get a problem
>>>>> with it.
>>>>
>>>>
>>>> Just tested.
>>>>
>>>> In this way it works.
>>>>
>>>> Is still a way to use a different Apache and Python ?
>>>
>>> Possibly just don't use Apache 2.2.11. Try with Apache 2.2.10 and
>>> whatever Python version you want.
>>>
>>> I am going to try and duplicate issue with Apache 2.2.11 tonight, so
>>> will see what I find.
>>
>> Duplicated problem.
>>
>> What is odd is that Apache 2.2.9 that comes with MacOS X is fine.
>>
>> A version of 2.2.9 that I had installed previously from source code also
>> worked.
>>
>> I then installed 2.2.11 and 2.2.10 and neither worked.
>>
>> I then rebuilt 2.2.9 from source again and it also then stopped working.
>>
>> When compiling from source code they are using APR that ships with
>> MacOS X which is an older version and probably shouldn't be used. Even
>> when I build Apache with option --with-included-apr, it still fails,
>> although the type of page that displays in Safari is different,
>> getting a blank screen rather than a page from Safari saying
>> connection had been dropped.
>>
>> Definitely something fishy going on, but have something to work with now.
>>
>> If this affects so many versions, don't understand why people haven't
>> been screaming about it.
>>
>> Can you both tell me what options you are supplying to 'configure'
>> when building Apache from source code.
>
> Possible fix/workaround (not sure which yet).
>
> Edit mod_wsgi.c and go to function Adapter_output_file().
>
> Search down for:
>
> b = apr_bucket_eos_create(r->connection->bucket_alloc);
> APR_BRIGADE_INSERT_TAIL(bb, b);
>
> Add in front of those two lines:
>
> /* XXX */
> b = apr_bucket_flush_create(r->connection->bucket_alloc);
> APR_BRIGADE_INSERT_TAIL(bb, b);
> /* XXX */
>
> Thus, you end up with:
>
> /* XXX */
> b = apr_bucket_flush_create(r->connection->bucket_alloc);
> APR_BRIGADE_INSERT_TAIL(bb, b);
> /* XXX */
>
> b = apr_bucket_eos_create(r->connection->bucket_alloc);
> APR_BRIGADE_INSERT_TAIL(bb, b);
>
> Need to work out why a flush is required as thought eos would achieve
> same result ans that has worked before.
Odd, leaving out both, ie.,
#if 0
/* XXX */
b = apr_bucket_flush_create(r->connection->bucket_alloc);
APR_BRIGADE_INSERT_TAIL(bb, b);
/* XXX */
b = apr_bucket_eos_create(r->connection->bucket_alloc);
APR_BRIGADE_INSERT_TAIL(bb, b);
#endif
also results in correct behaviour. Thus it is not so much the lack of
the flush, but that the eos is being inserted.
So, as fix while I work out what it should do, can also not add the
flush, but comment out the two eos lines.
Graham
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"modwsgi" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to
[email protected]
For more options, visit this group at
http://groups.google.com/group/modwsgi?hl=en
-~----------~----~----~----~------~----~------~--~---
