Failed test (32bits and slow computer)

[remicollet]

Hi,

First, on 64bits fast computer everything run fine :)

Trying to enable test suite in Fedora build, got some failure due to slow ARM (32bits) builder :

 There was 1 error:
 1) Predis\Connection\Aggregate\RedisClusterTest::testCanBeSerialized
 PHP_Invoker_TimeoutException: Execution aborted after 1 second
 /builddir/build/BUILDROOT/php-nrk-Predis-1.0.0-1.fc22.noarch/usr/share/tests/pear/Predis/Predis/Connection/Aggregate/RedisClusterTest.php:751
 --
 There were 3 failures:
 1) Predis\Command\HashIncrementByFloatTest::testIncrementsValueOfFieldByFloat
 Failed asserting that two strings are identical.
 --- Expected
 +++ Actual
 @@ @@
 -10.001
 +10.00099999999999945
 /builddir/build/BUILDROOT/php-nrk-Predis-1.0.0-1.fc22.noarch/usr/share/tests/pear/Predis/Predis/Command/HashIncrementByFloatTest.php:66
 2) Predis\Command\HashIncrementByFloatTest::testDecrementsValueOfFieldByFloat
 Failed asserting that two strings are identical.
 --- Expected
 +++ Actual
 @@ @@
 --10.001
 +-10.00099999999999945
 /builddir/build/BUILDROOT/php-nrk-Predis-1.0.0-1.fc22.noarch/usr/share/tests/pear/Predis/Predis/Command/HashIncrementByFloatTest.php:79
 3) Predis\Command\StringIncrementByFloatTest::testReturnsTheValueOfTheKeyAfterIncrement
 Failed asserting that '10.00000000000000178' matches expected 10.
 /builddir/build/BUILDROOT/php-nrk-Predis-1.0.0-1.fc22.noarch/usr/share/tests/pear/Predis/Predis/Command/StringIncrementByFloatTest.php:79
 FAILURES!
 Tests: 1904, Assertions: 4278, Failures: 3, Errors: 1.

For RedisClusterTest::testCanBeSerialized could it be possible to set this test with "@large" to avoid the timeout ?

For float failure, would it be possible to check rounded value (after 10 digits, p.e.) or to skip those tests if PHP_INT_SIZE<8 ?

[remicollet]

Full build.log https://kojipkgs.fedoraproject.org//work/tasks/4343/8014343/build.log (will be available for a short time)

[nrk]

Will fix in a couple of days. As for the float failure, I think I will go with the rounded value. Thanks!

[nrk]

Can you test if the tests run fine now? I purposely avoided to use round() everywhere, but since I can't reproduce the same output I can't verify. Also, I used @medium instead of @large as I think 10 seconds are more than enough for the serialization test even on very slow hardware.

[remicollet]

Test build https://kojipkgs.fedoraproject.org//work/tasks/3159/8063159/build.log

Everything seems ok now.
Thanks.

[remicollet]

"real" build (log won't be deleted)
http://koji.fedoraproject.org/koji/getfile?taskID=8063263&name=build.log

[nrk]

Thanks @remicollet!