On Tue, 1 Sep 2009 08:13:05 +0200, Kinkie <gkinkie_at_gmail.com> wrote:
> On Mon, Aug 31, 2009 at 5:49 PM, Kinkie<gkinkie_at_gmail.com> wrote:
>> bb:tweak
>>
>> This patch makes configure fail when aio is not available.
>>
>> Test-build on openbsd 4.5 fails with:
>>
>> checking for aio.h... no
>> checking for aio_read in -lrt... no
>> checking for aio_read in -laio... no
>> configure: error: POSIX AIO is not available on this system
>> buildtest.sh result is 1
>>
>> Do we really need at line 435:
>> AC_MSG_ERROR(POSIX AIO is not available on this system)) ? It stops the
>> build.
>>
>> Changing that line with these three:
>> with_aio=no
>> LIBAIO=""
>> AC_MSG_NOTICE(POSIX AIO is not available on this system))
>>
>> allows the build to complete successfully, and is probably more in
>> line with the intended purpose of this patch.
>
>
> Sorry, this change still doesn't fix the build.
> I'll try some more; vobsd is the perfect platform to test it as it
> doesn't have AIO.
Thanks kinkie. This kind of testing is exactly whats needed.
The enable-diskio=all and default loops need to be changed again from what
is in the patch. Unfortunately my evening has disappeared so I wont be
able to do that and help with the debug until too late probably.
Yes, the intention was to fail-soft when AIO was not available, but
fail-hard when something was _explicitly enabled_ that required it to work.
Amos
Received on Tue Sep 01 2009 - 06:23:41 MDT
This archive was generated by hypermail 2.2.0 : Tue Sep 01 2009 - 12:00:08 MDT