Обсуждение: Build broken on mips

Hi,

The build is broken on mips on the HEAD branch.
Here is a patch to fix that (builds and make check is OK on NetBSD/cobalt 7.99.1, GCC 4.8.4):


--- ../pgsql/src/include/storage/s_lock.h       2014-11-05 21:01:05.000000000 +0100
+++ src/include/storage/s_lock.h        2014-11-06 23:12:42.000000000 +0100
@@ -601,9 +601,10 @@               "       .set noreorder      \n" \               "       .set nomacro        \n" \
          "       sync                \n" \ 
-               "       .set pop              "
-:
-:              "memory");
+               "       .set pop              " \
+: \
+: \
+:              "memory"); \       *((volatile slock_t *) (lock)) = 0; \} while (0)



Regards,

Rémi Zara

On 11/08/2014 12:25 AM, Rémi Zara wrote:
> Hi,
>
> The build is broken on mips on the HEAD branch.
> Here is a patch to fix that (builds and make check is OK on NetBSD/cobalt 7.99.1, GCC 4.8.4):

Thanks, committed!

- Heikki

On 2014-11-07 23:25:50 +0100, Rémi Zara wrote:
> The build is broken on mips on the HEAD branch.
> Here is a patch to fix that (builds and make check is OK on
> NetBSD/cobalt 7.99.1, GCC 4.8.4):

Any chance you could put up a buildfarm animal (http://pgbuildfarm.org/)
up on that? Since nobody of the regular postgres developers has access
to mips that'd make it much more likely that we won't break it. And
generally make our lives easier.

Greetings,

Andres Freund

--
 Andres Freund                       http://www.2ndQuadrant.com/
 PostgreSQL Development, 24x7 Support, Training & Services

On 11/11/2014 01:08 AM, Heikki Linnakangas wrote:
> On 11/08/2014 12:25 AM, Rémi Zara wrote:
>> Hi,
>>
>> The build is broken on mips on the HEAD branch.
>> Here is a patch to fix that (builds and make check is OK on NetBSD/cobalt 7.99.1, GCC 4.8.4):
>
> Thanks, committed!

To verify this, I tried compiling PostgreSQL on a emulated qemu mips64el
Debian system, and this still doesn't work:

> gcc -O0 -Wall -Wmissing-prototypes -Wpointer-arith -Wdeclaration-after-statement -Wendif-labels
-Wmissing-format-attribute-Wformat-security -fno-strict-aliasing -fwrapv -fexcess-precision=standard -g
-I../../../../src/include-D_GNU_SOURCE   -c -o xlog.o xlog.c 
> xlog.c: In function ‘XLogInsertRecord’:
> xlog.c:1022:3: error: expected ‘(’ before ‘)’ token
> xlog.c: In function ‘ReserveXLogInsertLocation’:

Looks like I missed the additional colon you had in your patch. Oops.
I'll go fix that.

- Heikki

> Le 11 nov. 2014 à 00:16, Andres Freund <andres@2ndquadrant.com> a écrit :
>
> On 2014-11-07 23:25:50 +0100, Rémi Zara wrote:
>> The build is broken on mips on the HEAD branch.
>> Here is a patch to fix that (builds and make check is OK on
>> NetBSD/cobalt 7.99.1, GCC 4.8.4):
>
> Any chance you could put up a buildfarm animal (http://pgbuildfarm.org/)
> up on that? Since nobody of the regular postgres developers has access
> to mips that'd make it much more likely that we won't break it. And
> generally make our lives easier.
>

Hi,

I found the bug while reviving buildfarm member pika. You can see the results of some runs on the dashboard.

Thanks for committing a fix !

Regards,

Rémi

On Sun, Nov 16, 2014 at 01:02:59PM +0100, Rémi Zara wrote:
> > Le 11 nov. 2014 à 00:16, Andres Freund <andres@2ndquadrant.com> a écrit :
> > On 2014-11-07 23:25:50 +0100, Rémi Zara wrote:
> >> The build is broken on mips on the HEAD branch.
> >> Here is a patch to fix that (builds and make check is OK on
> >> NetBSD/cobalt 7.99.1, GCC 4.8.4):
> >
> > Any chance you could put up a buildfarm animal (http://pgbuildfarm.org/)
> > up on that? Since nobody of the regular postgres developers has access
> > to mips that'd make it much more likely that we won't break it. And
> > generally make our lives easier.
> >
>
> Hi,
>
> I found the bug while reviving buildfarm member pika. You can see the results of some runs on the dashboard.

Does it interest you to revive pika again?  It stopped reporting a few days
after your message.

Thanks,
nm

> Le 3 janv. 2015 à 21:57, Noah Misch <noah@leadboat.com> a écrit :
>
> On Sun, Nov 16, 2014 at 01:02:59PM +0100, Rémi Zara wrote:
>>> Le 11 nov. 2014 à 00:16, Andres Freund <andres@2ndquadrant.com> a écrit :
>>> On 2014-11-07 23:25:50 +0100, Rémi Zara wrote:
>>>> The build is broken on mips on the HEAD branch.
>>>> Here is a patch to fix that (builds and make check is OK on
>>>> NetBSD/cobalt 7.99.1, GCC 4.8.4):
>>>
>>> Any chance you could put up a buildfarm animal (http://pgbuildfarm.org/)
>>> up on that? Since nobody of the regular postgres developers has access
>>> to mips that'd make it much more likely that we won't break it. And
>>> generally make our lives easier.
>>>
>>
>> Hi,
>>
>> I found the bug while reviving buildfarm member pika. You can see the results of some runs on the dashboard.
>
> Does it interest you to revive pika again?  It stopped reporting a few days
> after your message.
>
> Thanks,
> nm


Hi,

Unfortunately, the operating system is not stable enough to provide meaningful results for a buildfarm member.
I will try again regularly with updated OS.

Regards,

Rémi