Investigate random php 5.6 failure in core branch tests.

As an update to this issue.

The Apache2 children are seg faulting

[Wed Jan 20 21:56:09.484749 2016] [core:notice] [pid 15] AH00051: child pid 3268 exit signal Segmentation fault (11), possible coredump in /etc/apache2
[Wed Jan 20 21:56:25.502063 2016] [core:notice] [pid 15] AH00051: child pid 3282 exit signal Segmentation fault (11), possible coredump in /etc/apache2
......
[Wed Jan 20 22:02:44.905457 2016] [core:notice] [pid 15] AH00051: child pid 3251 exit signal Segmentation fault (11), possible coredump in /etc/apache2

The failing test class that I've been focusing on is
Update.Drupal\system\Tests\Update\UpdatePathTestBaseFilledTest

This issue seems to mainly affect AWS. I've tried re-creating it on a virtualbox and actual hardware.

I think I've narrowed the issue down to APCu, though upgrading to the latest stable version that is compatible, 4.0.10, yielded the same intermittent issue.

I haven't been able to get a core dump file; I'm not sure if that is my inexperience with core dump files or the docker container.

[Sun Jan 24 05:59:10.173506 2016] [core:notice] [pid 6] AH00051: child pid 535 exit signal Segmentation fault (11), possible coredump in /var/apache2-gdb-dump
[Sun Jan 24 05:59:15.180937 2016] [core:notice] [pid 6] AH00051: child pid 649 exit signal Segmentation fault (11), possible coredump in /var/apache2-gdb-dump
[Sun Jan 24 05:59:25.192451 2016] [core:notice] [pid 6] AH00051: child pid 534 exit signal Segmentation fault (11), possible coredump in /var/apache2-gdb-dump
[Sun Jan 24 06:00:09.236549 2016] [core:notice] [pid 6] AH00051: child pid 652 exit signal Segmentation fault (11), possible coredump in /var/apache2-gdb-dump
zend_mm_heap corrupted

With APCu disabled the test runs fine. The tests seem to run just as fast with APCu turned off.

I haven't seen this segfault, but nearly every issue I've run into with APCu has usually been because it's got full/segmented and can't handle evictions - like the "can't allocate memory for pool' bug. Seems very likely that could happen with a full test run.

@catch I've been able to reproduce the error with running just one test group.

Seeing a lot of failures with D7 core update tests as well: https://www.drupal.org/pift-ci-job/189110

So does this always occur in Drupal\system\Tests\Update\UpdatePathTestBaseFilledTest or it is always happening in Drupal\system\Tests\Update\UpdatePathTestBaseTest which quite a few tests extend? Also it looks like the fail is happening here core/update.php/start?id=4&op=do_nojs - so this is where we build a graph of update dependencies - maybe this might give us a way to repeat this.

What's weird is that some similar random fails seem to go away if ConfigDependencyManager.php is patched in anyway... see #2762549: Drupal\field\Tests\Update\FieldUpdateTest, Drupal\views\Tests\Update\EntityViewsDataUpdateTest and Drupal\comment\Tests\CommentFieldsTest fail on 8.1.x

Facets has similar random test fails for php5.6 as well, there are no failures on 5.5 or 7.0. https://www.drupal.org/pift-ci-job/369229

Drupal\Tests\facets\FunctionalJavascript\WidgetJSTest::testAddFacet
PHPUnit_Framework_Exception: Segmentation fault (core dumped)

I haven't looked into this yet, but I figured an extra datapoint might be useful.