204894 2019-12-05 09:26:42 -0800 EWS: JSC-Tests-EWS should run with debug instead of release 2020-03-21 06:17:52 -0700 1 1 1 Unclassified WebKit Tools / Tests WebKit Nightly Build Unspecified Unspecified RESOLVED WONTFIX InRadar P2 Normal --- 205279 1 jbedard jbedard aakash_jain ap dean_johnson fpizlo mark.lam webkit-bug-importer oldest_to_newest 1595840 0 jbedard 2019-12-05 09:26:42 -0800 This is a result of a conversation with Mark, who indicated that a debug EWS is probably more valuable than then a release one. 1595846 1 384912 jbedard 2019-12-05 09:34:56 -0800 Created attachment 384912 Patch 1595847 2 384912 ap 2019-12-05 09:36:03 -0800 Comment on attachment 384912 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=384912&action=review > Tools/ChangeLog:3 > + EWS: JSC-Tests-EWS should run with debug instead of release I'm somewhat skeptical, because historically issues on debug JSC tests were often discarded or deprioritized. Are we seeing many debug-only regressions on JSC tests that are getting fixed? 1595848 3 ap 2019-12-05 09:36:10 -0800 <rdar://problem/57643755> 1595849 4 mark.lam 2019-12-05 09:40:35 -0800 I found out last night that the debug build runs ridiculously slow. Let me do something to make it complete in a more reasonable time before we apply this change. (In reply to Alexey Proskuryakov from comment #2) > Comment on attachment 384912 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=384912&action=review > > > Tools/ChangeLog:3 > > + EWS: JSC-Tests-EWS should run with debug instead of release > > I'm somewhat skeptical, because historically issues on debug JSC tests were > often discarded or deprioritized. > > Are we seeing many debug-only regressions on JSC tests that are getting > fixed? With a debug build, we check for missing exception checks, and various GC bugs. These bugs will go unnoticed if we don't run with the debug build. That said, let me put together a debug build configuration that makes more sense to run in the EWS first. Stay tuned. 1595856 5 ap 2019-12-05 09:54:12 -0800 If there is one that the JSC team is willing to fully defend, we should run it in post-commit testing too, and unskip tests that are marked as skipped in debug. 1598930 6 mark.lam 2019-12-16 09:37:49 -0800 After https://bugs.webkit.org/show_bug.cgi?id=205279 lands, there's one more change that needs to be made in order to have the tests complete in a reasonable amount of time: for the EWS, I recommend only running the default configuration of the tests. That will get us a lot of mileage. To only run the default configuration, you can pass --filter "\.default" to the run-javascriptcore-tests command line in addition to the --debug flag. 1598932 7 384912 mark.lam 2019-12-16 09:39:01 -0800 Comment on attachment 384912 Patch Please also add the command line option --filter '\.default'. 1598972 8 384912 fpizlo 2019-12-16 10:45:23 -0800 Comment on attachment 384912 Patch This isn't a great idea. Debug testing takes longer, experiences races that are different from the release ones, and has important behavioral changes (object layout, synchronization) that make it quite different from release. If we test one thing it should be release. If we test two things it should be two different kinds of release builds. 1599483 9 jbedard 2019-12-17 11:50:39 -0800 After talking with various JSC folks, seems like this is a 'WON'T FIX', although we might modify release builds to do more debug-y things, that seems like it should be another bug. 384912 2019-12-05 09:34:56 -0800 2019-12-16 09:39:01 -0800 Patch bug-204894-20191205093456.patch text/plain 1174 jbedard SW5kZXg6IFRvb2xzL0NoYW5nZUxvZwo9PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBUb29scy9DaGFuZ2VMb2cJKHJl dmlzaW9uIDI1MzE1MikKKysrIFRvb2xzL0NoYW5nZUxvZwkod29ya2luZyBjb3B5KQpAQCAtMSwz ICsxLDEyIEBACisyMDE5LTEyLTA1ICBKb25hdGhhbiBCZWRhcmQgIDxqYmVkYXJkQGFwcGxlLmNv bT4KKworICAgICAgICBFV1M6IEpTQy1UZXN0cy1FV1Mgc2hvdWxkIHJ1biB3aXRoIGRlYnVnIGlu c3RlYWQgb2YgcmVsZWFzZQorICAgICAgICBodHRwczovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1 Zy5jZ2k/aWQ9MjA0ODk0CisKKyAgICAgICAgUmV2aWV3ZWQgYnkgTk9CT0RZIChPT1BTISkuCisK KyAgICAgICAgKiBCdWlsZFNsYXZlU3VwcG9ydC9ld3MtYnVpbGQvY29uZmlnLmpzb246CisKIDIw MTktMTItMDUgIHlvdWVubiBmYWJsZXQgIDx5b3Vlbm5AYXBwbGUuY29tPgogCiAgICAgICAgIGlu c3BlY3Rvci9wYWdlL292ZXJyaWRlU2V0dGluZy1Nb2NrQ2FwdHVyZURldmljZXNFbmFibGVkLmh0 bWwgaXMgZmFpbGluZyBhZnRlciByZW1vdmFsIG9mIGludGVybmFscy5zZXRNb2NrTWVkaWFDYXB0 dXJlRGV2aWNlc0VuYWJsZWQgQVBJCkluZGV4OiBUb29scy9CdWlsZFNsYXZlU3VwcG9ydC9ld3Mt YnVpbGQvY29uZmlnLmpzb24KPT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gVG9vbHMvQnVpbGRTbGF2ZVN1cHBvcnQv ZXdzLWJ1aWxkL2NvbmZpZy5qc29uCShyZXZpc2lvbiAyNTMxNTEpCisrKyBUb29scy9CdWlsZFNs YXZlU3VwcG9ydC9ld3MtYnVpbGQvY29uZmlnLmpzb24JKHdvcmtpbmcgY29weSkKQEAgLTQzNiw3 ICs0MzYsNyBAQAogICAgICAgImljb24iOiAiYnVpbGRBbmRUZXN0IiwKICAgICAgICJmYWN0b3J5 IjogIkpTQ1Rlc3RzRmFjdG9yeSIsCiAgICAgICAicGxhdGZvcm0iOiAibWFjLWhpZ2hzaWVycmEi LAotICAgICAgImNvbmZpZ3VyYXRpb24iOiAicmVsZWFzZSIsCisgICAgICAiY29uZmlndXJhdGlv biI6ICJkZWJ1ZyIsCiAgICAgICAicnVuVGVzdHMiOiAidHJ1ZSIsCiAgICAgICAid29ya2VybmFt ZXMiOiBbImV3czEyNyIsICJld3MxMjgiXQogICAgIH0sCg==