73326 2011-11-29 08:22:00 -0800 REGRESSION (r100471-r100492): build-webkittestrunner fails on Mac bots 2011-11-30 13:58:41 -0800 1 1 1 Unclassified WebKit Tools / Tests 528+ (Nightly build) Unspecified Unspecified RESOLVED FIXED MakingBotsRed, PlatformOnly, Regression P2 Normal --- 1 aroben webkit-unassigned abarth cfleizach dpranke eric ojan sam oldest_to_newest 510786 0 aroben 2011-11-29 08:22:00 -0800 build-webkittestrunner has been failing on the Mac bots since somewhere between r100471 and r100492. build-webkittestrunner succeeded and tests ran: http://build.webkit.org/builders/SnowLeopard%20Intel%20Release%20%28WebKit2%20Tests%29/builds/15949 build-webkittestrunner failed and no tests ran: http://build.webkit.org/builders/SnowLeopard%20Intel%20Release%20%28WebKit2%20Tests%29/builds/15950 510787 1 aroben 2011-11-29 08:22:55 -0800 r100488 seems the most likely culprit: http://trac.webkit.org/changeset/100488 510789 2 aroben 2011-11-29 08:23:15 -0800 It's too bad that NRWT doesn't show us the build log! 510792 3 eric 2011-11-29 08:26:33 -0800 Well, that's intentional. :) But I agree, it should show us the build log in verbose mode (which the bots use). 510799 4 aroben 2011-11-29 08:29:13 -0800 The Debug bot shows the same thing: r100471: good http://build.webkit.org/builders/SnowLeopard%20Intel%20Debug%20%28WebKit2%20Tests%29/builds/2480 r100492: bad http://build.webkit.org/builders/SnowLeopard%20Intel%20Debug%20%28WebKit2%20Tests%29/builds/2481 510800 5 cfleizach 2011-11-29 08:29:52 -0800 Are there build failure logs available anywhere? (In reply to comment #1) > r100488 seems the most likely culprit: > > http://trac.webkit.org/changeset/100488 510801 6 eric 2011-11-29 08:30:54 -0800 No. See bug 71160. 510947 7 aroben 2011-11-29 11:31:39 -0800 Looks like Lion Intel Debug (WebKit2 Tests) is fine. Maybe there's a compiler error that affects GCC but not clang? 511766 8 aroben 2011-11-30 08:19:26 -0800 (In reply to comment #7) > Looks like Lion Intel Debug (WebKit2 Tests) is fine. Maybe there's a compiler error that affects GCC but not clang? Lion Intel Release (WebKit2 Tests) is now using Clang (thanks to Xcode 4.2), and it's still failing: http://build.webkit.org/builders/Lion%20Intel%20Release%20%28WebKit2%20Tests%29/builds/2136/steps/layout-test/logs/stdio 511767 9 aroben 2011-11-30 08:25:11 -0800 Looks like Lion Intel Debug (WebKit2 Tests) is now failing too. I guess the newer Clang in Xcode 4.2 is being more strict. 512001 10 aroben 2011-11-30 12:20:26 -0800 Hooray, we have build output! http://build.webkit.org/builders/Lion%20Intel%20Release%20%28WebKit2%20Tests%29/builds/2144/steps/layout-test/logs/stdio 512008 11 aroben 2011-11-30 12:33:58 -0800 Here's more: http://build.webkit.org/builders/Lion%20Intel%20Debug%20%28WebKit2%20Tests%29/builds/2062/steps/layout-test/logs/stdio 512009 12 aroben 2011-11-30 12:34:19 -0800 Both of those logs look like the problem is local to the bots. A clean build should fix it. 512021 13 aroben 2011-11-30 12:39:49 -0800 I've deleted the WebKitBuild directories on the two affected bots. Let's see what happens. 512056 14 aroben 2011-11-30 13:07:46 -0800 Looks like a clean build didn't fix it: http://build.webkit.org/builders/Lion%20Intel%20Release%20%28WebKit2%20Tests%29/builds/2146/steps/layout-test/logs/stdio 512061 15 aroben 2011-11-30 13:11:44 -0800 I'll try deleting precompiled headers on these machines too. 512105 16 aroben 2011-11-30 13:55:35 -0800 That seems to have done the trick: http://build.webkit.org/builders/Lion%20Intel%20Release%20%28WebKit2%20Tests%29/builds/2147/steps/layout-test/logs/stdio I'll wait for the other bot before closing this. 512113 17 aroben 2011-11-30 13:58:41 -0800 The other bot looks good too: http://build.webkit.org/builders/Lion%20Intel%20Debug%20%28WebKit2%20Tests%29/builds/2065/steps/layout-test/logs/stdio Yay!