REGRESSION: LayoutTest js/dom/imul.html is a flaky failure

• Component: JavaScriptCore
• Status: RESOLVED CONFIGURATION CHANGED

[Ryan Haddad]

LayoutTest js/dom/imul.html is a flaky failure

https://build.webkit.org/results/Apple%20El%20Capitan%20Release%20WK2%20(Tests)/r212130%20(13087)/results.html

https://webkit-test-results.webkit.org/dashboards/flakiness_dashboard.html#showAllRuns=true&tests=js%2Fdom%2Fimul.html

--- /Volumes/Data/slave/elcapitan-release-tests-wk2/build/layout-test-results/js/dom/imul-expected.txt
+++ /Volumes/Data/slave/elcapitan-release-tests-wk2/build/layout-test-results/js/dom/imul-actual.txt
@@ -44,7 +44,7 @@
 PASS testIMul(2.5,-2.5,10000) is -40000
 PASS testIMul(NaN, 1, 10000) is 0
 PASS testIMul(Infinity, 1, 10000) is 0
-PASS testIMul(1e40, 1, 10000) is 0
+FAIL testIMul(1e40, 1, 10000) should be 0. Was 3108634624.
 PASS testIMul(1, NaN, 10000) is 0
 PASS testIMul(1, Infinity, 10000) is 0
 PASS testIMul(1, 1e40, 10000) is 0

[Ryan Haddad]

First failure on the flakiness dashboard was 2/9/2017 12:00:10 PM @ r211973

[Ryan Haddad]

There is no obviously related change (that I can see, at least).

[Alexey Proskuryakov]

Looks like this happened once or twice, and isn't happening any more. The latest failure so far was in r212169.

Very suspicious that this affected multiple the bots at once, and quite concerning that it's just basic math that is failing.

Filip, does this ring a bell? Could this be made more reproducible with any debug flags?

[Alexey Proskuryakov]

Checking other testers, I see failures on February 6th and on 14th. This is a very rare failure, but definitely a recent regression, as it hasn't been seen for at least 3 months before.

[Radar WebKit Bug Importer]

<rdar://problem/30529735>

[Alexey Proskuryakov]

The diff is always the same BTW, with 3108634624 (0xB94A0000) instead of 0.

[Alexey Proskuryakov]

Not happening any more on bots, the latest occurrence was on April 15th.

[Filip Pizlo]

Resolved since mid-Apr as ap says.