On Sat, May 25 2019, Charlene Wendling <[email protected]> wrote:
> Hi,
>
> Thanks again for taking care of it!
>
>> On Tue, May 14 2019, Jeremie Courreges-Anglas <[email protected]> wrote:
>> > On Mon, May 13 2019, Jeremie Courreges-Anglas <[email protected]>
>> > wrote:
>> >> Now that 10.2 has been committed, here's an update to the latest.
>> >> Upstream removed the example config file so I decided to create one
>> >> based on
>> >>
>> >>   https://mariadb.com/kb/en/library/configuring-mariadb-with-option-files/
>> >>
>> >> No symbol change in the SHARED_LIBS.  The diff also includes a fix
>> >> for the test target (mysql-test tests don't finish here).
>> >>
>> >> Looking for real use reports and oks.
>> >
>> > no_server doesn't build any more because my_global.h file is
>> > missing, so I'm holding off for now.  Test reports still welcome.
>> 
>> bah, I'm not sure why breaking the no_server PSEUDO_FLAVOR would
>> matter. The problem is known upstream and I worked around it in the
>> diff below.
>> 
>>   https://jira.mariadb.org/browse/MDEV-19152
>> 
>> Meanwhile, cwen@ reported that mariadb-10.2.23 doesn't build on
>> powerpc because it's lacking __sync_* atomics.  10.3.x switched to
>> __atomic_* builtins which are better supported by gcc-8.3.0, this
>> fixes mariadb on powerpc according to cwen's tests (thanks!).
>
> I've retested your diff and met no issues at all. 
>
> Please note that it also needs the infamous atomics block - they
> provide proper atomics detection... but no -latomic flag test.

sigh, I thought it was working out of the box, but it looks like very
few projects get this -latomic part right...

> I'm sending a diff of the Makefile, but you know very well what i'm
> talking about ;)

Yep, I see.

I have committed this, thanks.

-- 
jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF  DDCC 0DFA 74AE 1524 E7EE

Reply via email to