[Webkit-unassigned] [Bug 202011] New: stress/test-out-of-memory.js is not throwing OOM into ARMv7 and MIPS

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Thu Sep 19 15:44:40 PDT 2019


https://bugs.webkit.org/show_bug.cgi?id=202011

            Bug ID: 202011
           Summary: stress/test-out-of-memory.js is not throwing OOM into
                    ARMv7 and MIPS
           Product: WebKit
           Version: WebKit Nightly Build
          Hardware: Unspecified
                OS: Unspecified
            Status: NEW
          Severity: Normal
          Priority: P2
         Component: JavaScriptCore
          Assignee: webkit-unassigned at lists.webkit.org
          Reporter: ticaiolima at gmail.com

After https://trac.webkit.org/r249578, the test `stress/test-out-of-memory.js` was introduced to test some paths that are taken during memory allocation, mainly the path where `allocationLarge` fails due to unavailable space for large allocations (m_largeFree). However, this test has some assumptions that are not valid into ARMv7 and MIPS ports. The current behavior of the test in those architectures is that it does not throw during `new ArrayBuffer(1000)` allocation site, because eden collection keeps happening between iterations. The collection is trigged into those architectures because the amount of stress `new Promise` generates into GC limits is not enough to avoid collection while loop is executing.
Changing the size of `UInt8Array` from `80000000` to `160000000` enables us to finally avoid collection happening during `ArrayBuffer` allocation loop, but we can't guarantee this test is always going to execute without error when Gigacage is disabled, given we can reach an OOM state in some allocations that need to succeed, turning this test flawky for those architectures.
Given that, I think we should skip this test for ARMv7 and MIPS.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20190919/fefdc84e/attachment-0001.html>


More information about the webkit-unassigned mailing list