On Sat, Nov 23, 2013 at 3:45 PM, Jim Jagielski <jim@jagunet.com> wrote:
I'm just curious why people never saw this on trunk...
That code path is almost exactly a year old. No one has
hit this in 12 months??

l rarely try trunk except on

* Windows, which uses a system mechanism regardless of compiler
* Linux or FreeBSD, 64-bit gcc builds

--/--

I'm curious what other scenarios will fail though.  I can try Sun Studio on Solaris 10 x86_64 (32-bit and 64-bit builds) "soon".  But Sun Studio on SPARC presumably uses different explicit code in APR and I don't have access to that.

Right now it looks reasonable to change the ia32.c and os390/atomic.c implementations in APR as described before, but there's no substitute for actually trying the other atomic implementations, which look safe to me but with little certainty.



On Nov 23, 2013, at 12:45 PM, Rainer Jung <rainer.jung@kippdata.de> wrote:

> On 23.11.2013 14:15, Jeff Trawick wrote:
>> On Sat, Nov 23, 2013 at 8:04 AM, Rainer Jung <rainer.jung@kippdata.de
>> <mailto:rainer.jung@kippdata.de>> wrote:
>>
>>    On 22.11.2013 23:03, Jim Jagielski wrote:
>>> Anyone ever try OpenPA?
>>>
>>>      https://trac.mcs.anl.gov/projects/openpa/
>>>
>>> It's under MIT, fwiw.
>>
>>    Haven't tried it but the README
>>
>>    http://git.mcs.anl.gov/radix/openpa.git/blob_plain/HEAD:/README
>>
>>    indicates only platform support based on gcc plus Windows (but they say
>>    they haven't set up a build system for it) plus Solaris (but that seems
>>    to be Solaris 10+ only, at least my oldest Solaris 8 doesn't seem to
>>    have the needed implementation).
>>
>>    Regards,
>>
>>    Rainer
>>
>>
>> By the way, Rainer, in your l*m*n test combinations, which compiler(s)
>> did you use on Solaris while confirming that 2.4.7 Event works at least
>> a little?
>
> I'm only building with gcc. On Solaris 8 with gcc 4.1.2, on Solaris 10
> the latest and greatest (4.8.2).
>
> Will post the full 2.4.7 results a bit later (only for the sake of
> completeness), but the builds are also broken for event on Linux 32 Bits
> even when using gcc:
>
> [Thu Nov 21 11:34:53.149765 2013] [mpm_event:crit] [pid 18492:tid
> 3080439472] AH02405: atomics not working as expected
>
> Sorry for not having checked progress during the tests, but the last
> combinations just finished right now.
>
> Regards,
>
> Rainer
>




--
Born in Roswell... married an alien...
http://emptyhammock.com/