Bug 149900

Summary: Disable tail calls because it is breaking some sites.
Product: WebKit Reporter: Mark Lam <mark.lam>
Component: JavaScriptCoreAssignee: Mark Lam <mark.lam>
Status: RESOLVED FIXED    
Severity: Normal CC: basile_clement, msaboff, ossy, saam, zan
Priority: P2    
Version: WebKit Local Build   
Hardware: Unspecified   
OS: Unspecified   
Attachments:
Description Flags
the patch.
saam: review+
follow up patch.
saam: review+
follow up 2: need to suppress JSC tail call tests none

Mark Lam
Reported 2015-10-07 14:59:55 PDT
This is until we fix whatever the breakage is.
Attachments
the patch. (1.30 KB, patch)
2015-10-07 15:05 PDT, Mark Lam
saam: review+
follow up patch. (2.13 KB, patch)
2015-10-07 16:31 PDT, Mark Lam
saam: review+
follow up 2: need to suppress JSC tail call tests (4.36 KB, patch)
2015-10-07 17:31 PDT, Mark Lam
no flags
Mark Lam
Comment 1 2015-10-07 15:05:31 PDT
Created attachment 262647 [details] the patch.
Saam Barati
Comment 2 2015-10-07 15:06:17 PDT
Comment on attachment 262647 [details] the patch. r=me
Mark Lam
Comment 3 2015-10-07 15:10:28 PDT
Mark Lam
Comment 4 2015-10-07 16:30:53 PDT
Need to update expected test results to match this change.
Mark Lam
Comment 5 2015-10-07 16:31:23 PDT
Created attachment 262654 [details] follow up patch.
Mark Lam
Comment 6 2015-10-07 16:36:58 PDT
Thanks. Follow up patch landed in r190695: <http://trac.webkit.org/r190695>.
Mark Lam
Comment 7 2015-10-07 17:31:48 PDT
Created attachment 262662 [details] follow up 2: need to suppress JSC tail call tests This should be the last follow up (based on all the tests in http://trac.webkit.org/changeset/190289). This follow up 2 patch has been rubber stamped by Saam. I will land it as soon as my test run confirms that I've addressed the remaining issue.
Mark Lam
Comment 8 2015-10-07 17:49:25 PDT
JSC tests are now passing. Follow up 2 landed in r190699: <http://trac.webkit.org/r190699>.
Alexey Proskuryakov
Comment 9 2015-10-07 21:59:37 PDT
Looks like follow up 2 changed http/tests/misc/large-js-program.php from crash to timeout on Windows Release. Could you please add an appropriate expectation or skip? https://webkit-test-results.webkit.org/dashboards/flakiness_dashboard.html#showAllRuns=true&tests=http%2Ftests%2Fmisc%2Flarge-js-program.php
Alexey Proskuryakov
Comment 10 2015-10-07 22:00:47 PDT
Mark Lam
Comment 11 2015-10-07 22:10:50 PDT
(In reply to comment #10) > Also, tail call tests apparently need to be disabled: > https://build.webkit.org/builders/Apple%20Win%207%20Debug%20%28Tests%29/ > builds/67893/steps/jscore-test/logs/stdio They are already. Build 67893 was built with r190692. The tests were skipped in r190699.
Mark Lam
Comment 12 2015-10-07 22:41:40 PDT
(In reply to comment #9) > Looks like follow up 2 changed http/tests/misc/large-js-program.php from > crash to timeout on Windows Release. Could you please add an appropriate > expectation or skip? > > https://webkit-test-results.webkit.org/dashboards/flakiness_dashboard. > html#showAllRuns=true&tests=http%2Ftests%2Fmisc%2Flarge-js-program.php The flakiness dashboard says that the blame list for that result change is https://trac.webkit.org/log/?verbose=on&rev=190700&stop_rev=190698. Of those revisions, r190699 is for follow up 2. But r190699 only changes the JSC tests. It should have no effect on the layout tests. On the other hand, r190698 is about changing the testing infrastructure for Windows. It is more likely that r190698 triggered the change in test result.
Zan Dobersek
Comment 13 2015-10-13 00:37:55 PDT
Before they were disabled, I was experiencing crashes due to tail call optimizations on ARM Thumb and Linux. Filed bug #150083 with some information.
Note You need to log in before you can comment on or make changes to this bug.