httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jim Jagielski <...@jaguNET.com>
Subject Re: eventopt observations
Date Thu, 20 Jun 2013 11:56:41 GMT
Thx for the feedback! Comments inline.

On Jun 19, 2013, at 4:16 PM, Rainer Jung <rainer.jung@kippdata.de> wrote:

> While testing the new APR 1.4.8 I ran the httpd test suite on trunk, the
> first time since long.
> 
> 1) eventopt didn't run on SLES 10 32 Bit
> ========================================
> 
> Error message:
> 
> [Wed Jun 19 14:24:21.443541 2013] [mpm_eventopt:crit] [pid 25540:tid
> 3082651312] AH02406: atomics not working as expected
> 
> http://svn.apache.org/viewvc?view=revision&revision=r1410459
> 
> I don't really understand why the test is supposed to work, it
> decrements below "0". The test also fails for the event MPM, but hans't
> (yet) been backported to 2.4.
> 

This is related to fdqueue.c:

apr_status_t ap_queue_info_set_idle(fd_queue_info_t * queue_info,
                                    apr_pool_t * pool_to_recycle)
{
    apr_status_t rv;
    int prev_idlers;

    ap_push_pool(queue_info, pool_to_recycle);

    /* Atomically increment the count of idle workers */
    /*
     * TODO: The atomics expect unsigned whereas we're using signed.
     *       Need to double check that they work as expected or else
     *       rework how we determine blocked.
     * UPDATE: Correct operation is performed during open_logs()
     */


> 2) include.t
> ============
> 
> I noticed sporadic test failures for include.t when checking the file
> time. The time zone expected by the test framework was CET, but the time
> zone returned by mod_include was UTC.
> 
> I noticed it only happens for the eventopt MPM and checked the code. I
> found the function force_set_tz() that sets the TZ variable to UTC if it
> is not already set.
> 
> At this point I only wanted to note that this breaks include.t:
> 
> # Failed test 59 in t/modules/include.t at line 349

I am guessing that a simple fix would be to adjust the test framework
to use UTC... I did do some benchmarking and confirmed Paul's
observation regarding the timezone issue.

Mime
View raw message