Changeset 128391 in webkit


Ignore:
Timestamp:
Sep 12, 2012 5:49:47 PM (12 years ago)
Author:
Stephanie Lewis
Message:

Reduce parallism on the wk2 testers.
https://bugs.webkit.org/show_bug.cgi?id=95906

Reviewed by Dirk Pranke.

The wk2 testers on Mountain Lion are getting stuck and timing out. The problem appears to be due to
resource contention. Reducing the number of processes alleviates the issue.

Starting by reducing 25%.

  • Scripts/webkitpy/layout_tests/port/mac.py:

(MacPort.default_child_processes):

Location:
trunk/Tools
Files:
2 edited

Legend:

Unmodified
Added
Removed
  • trunk/Tools/ChangeLog

    r128388 r128391  
     12012-09-12  Stephanie Lewis  <slewis@apple.com>
     2
     3        Reduce parallism on the wk2 testers.
     4        https://bugs.webkit.org/show_bug.cgi?id=95906
     5
     6        Reviewed by Dirk Pranke.
     7
     8        The wk2 testers on Mountain Lion are getting stuck and timing out.  The problem appears to be due to
     9        resource contention.  Reducing the number of processes alleviates the issue. 
     10
     11        Starting by reducing 25%.
     12
     13        * Scripts/webkitpy/layout_tests/port/mac.py:
     14        (MacPort.default_child_processes):
     15
    1162012-09-12  Brady Eidson  <beidson@apple.com>
    217
  • trunk/Tools/Scripts/webkitpy/layout_tests/port/mac.py

    r126533 r128391  
    116116        default_count = super(MacPort, self).default_child_processes()
    117117
     118        # FIXME: https://bugs.webkit.org/show_bug.cgi?id=95906  With too many WebProcess WK2 tests get stuck in resource contention.
     119        # To alleviate the issue reduce the number of running processes
     120        # Anecdotal evidence suggests that a 4 core/8 core logical machine may run into this, but that a 2 core/4 core logical machine does not.
     121        if self.get_option('webkit_test_runner') and default_count > 4:
     122            default_count = int(.75 * default_count)
     123
    118124        # Make sure we have enough ram to support that many instances:
    119125        total_memory = self.host.platform.total_bytes_memory()
Note: See TracChangeset for help on using the changeset viewer.