Mailing-List: contact dev-help@apr.apache.org; run by ezmlm
Precedence: bulk
Date: Fri, 02 Jul 2004 18:24:38 -0700
From: Justin Erenkrantz <justin@erenkrantz.com>
To: dev@apr.apache.org
Subject: Re: FreeBSD APR patch for threads
Message-ID: <0AFC542D376F0D404528B0FB@st-augustin.ics.uci.edu>
In-Reply-To: <83AE97E7-CC65-11D8-8973-000393B3C494@clove.org>
References: <20040702194354.GA60584@crodrigues.org>
 <83AE97E7-CC65-11D8-8973-000393B3C494@clove.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

--On Friday, July 2, 2004 1:22 PM -0700 Aaron Bannert <aaron@clove.org> 
wrote:

>> This patch basically does the following:
>>  - for the latest versions of FreeBSD, puts in the correct flags for
>>    linking in threads
>>  - for older versions of FreeBSD, i.e FreeBSD 4.x, disables threads by
>>    default (which is what APR does right now)
>
> I'm +1 for this patch in concept, but I don't have a FreeBSD machine
> with an osreldate recent enough to test this patch.

As noted in httpd's STATUS, FreeBSD 5.2 isn't sufficient as the default 
thread library there (libc_r) still breaks with httpd's worker.  I've also 
seen breakage with Subversion if threads are enabled: the commit for SSL 
cert prompts will not display correctly.  If you play with libmap.conf to 
use libkse, httpd will serve pages correctly with 5.2.  The good news is 
that 5.3 (and -CURRENT now, I believe) will switch to libkse by default...

That said, I'm not quite clear what the semantics of kern.osreldate are - 
will that block off any 5.2.x (i.e. 5.2.2 if that is released) and only 
allow current -CURRENT and eventually 5.3 and beyond?  My 5.2.1 boxes 
report '502010', which is less than '502102' - so it would prevent them 
from enabling threads.  I just want to make sure that it doesn't trigger on 
a 5.2.2 which, AIUI, will still use libc_r.  If it does block 5.2.x, then 
this patch gets +1.  Regardless, this is not something that must go into 
APR 1.0 - we can commit it after 1.0 (which should have been today!).  -- 
justin