++1
> On Sep 25, 2015, at 7:07 AM, Plüm, Rüdiger, Vodafone Group
> <[email protected]> wrote:
>
>
>
>> -----Ursprüngliche Nachricht-----
>> Von: Joe Orton [mailto:[email protected]]
>> Gesendet: Freitag, 25. September 2015 12:47
>> An: [email protected]
>> Betreff: Re: logio problem with SSL
>>
>> On Fri, Sep 25, 2015 at 09:50:04AM +0200, Yann Ylavic wrote:
>>> On Fri, Sep 25, 2015 at 1:00 AM, Yann Ylavic <[email protected]>
>> wrote:
>>>> On Fri, Sep 25, 2015 at 12:22 AM, Eric Covener <[email protected]>
>> wrote:
>>>>>
>>>>> two logs (http/https) sorted to top of autoindex here:
>>>>> http://people.apache.org/~covener/
>>>>
>>>> Looks like mod_ssl should also forward EOR buckets.
>>>>
>>>> Does this work:
>>>> Index: modules/ssl/ssl_engine_io.c
>>>> ===================================================================
>>>> --- modules/ssl/ssl_engine_io.c (revision 1705160)
>>>> +++ modules/ssl/ssl_engine_io.c (working copy)
>>>> @@ -1707,12 +1707,12 @@ static apr_status_t
>> ssl_io_filter_output(ap_filter
>>>
>>> I committed this one in r1705194, and also the one preventing the
>>> FLUSH for non-blocking bio_filter_in_read() in r1705236.
>>> You may not want to apply the latter, for your testing path to be
>>> consistent with what you had so far...
>>
>> The behaviour of that loop is quite bad, it will treat a single brigade
>> like <EOS EOC> differently to two separate brigades <EOS> <EOC>,
>> although that should never happen in practice... currently.
>>
>> I'm not sure what the "correct" behaviour of connection-level filters
>> should be with metadata buckets. I could argue they should delete
>> everything they don't understand. mod_ssl should not care at all about
>> EOS or EOR.
>>
>> But dodging that issue... simplifying the loop like this, does that
>> still work for the logio issue?
>
> Haven't looked at the logio issue, but this makes sense. +1. The previous
> different handling
> of the buckets depending on whether they are split across two brigades or
> just in one seems wrong.
>
> Regards
>
> Rüdiger
>