205299 2019-12-16 14:13:35 -0800 REGRESSION: animations/leak-document-with-css-animation.html is very flaky 2019-12-24 04:39:23 -0800 1 1 1 Unclassified WebKit Animations WebKit Local Build Unspecified Unspecified RESOLVED FIXED InRadar P2 Normal --- 1 ap graouts commit-queue dino graouts koivisto webkit-bug-importer oldest_to_newest 1599094 0 ap 2019-12-16 14:13:35 -0800 animations/leak-document-with-css-animation.html only failed once between 2019-10-10 and 2019-11-14, and now it's failing multiple times each day. https://results.webkit.org/?suite=layout-tests&test=animations%2Fleak-document-with-css-animation.html @@ -5,7 +5,7 @@ The iframe has finished loading. The iframe has been destroyed. -PASS internals.numberOfLiveDocuments() is numberOfLiveDocumentsAfterIframeLoaded - 1 +FAIL internals.numberOfLiveDocuments() should be 7. Was 6. PASS successfullyParsed is true 1599095 1 webkit-bug-importer 2019-12-16 14:13:46 -0800 <rdar://problem/57984602> 1599097 2 ap 2019-12-16 14:16:06 -0800 Marked as flaky in http://trac.webkit.org/r253575 1599339 3 graouts 2019-12-17 02:09:20 -0800 I made two changes in Web Animations on November 14: r252455 and r252461. 1599343 4 graouts 2019-12-17 02:14:32 -0800 Of note, there was a failure with a run with r252460: https://build.webkit.org/results/Apple%20High%20Sierra%20Release%20WK1%20(Tests)/r252460%20(17046)/animations/leak-document-with-css-animation-pretty-diff.html 1599720 5 graouts 2019-12-18 02:08:33 -0800 Running many iterations of the test itself doesn't reveal any issues. Need to try running tests in the same order of a worker that showed a failure next. 1599721 6 graouts 2019-12-18 02:18:40 -0800 No luck with the test list from https://build.webkit.org/builders/Apple-Catalina-Release-WK2-Tests/builds/1595. 1599723 7 graouts 2019-12-18 02:27:17 -0800 Something that puzzles me is that when the test runs in isolation, it has 2 live documents after the iframe has loaded, and 1 after it has been destroyed. However, when it fails on the bots it shows that the number of live documents is much higher. I suspect that internals.numberOfLiveDocuments() returns the number of live documents across runners. At any rate, the figures reported shows a lower number of live documents still, which is good. So maybe we ought to rework this test to not check for a hard number, but only a lower number after the iframe has been destroyed. 1599728 8 graouts 2019-12-18 02:59:27 -0800 Actually, I think we should call `gc()` before we take the reference snapshot of live documents, hopefully this will fix the flakiness by allowing any document pending destruction to be destroyed. If flakiness persists, I think we can then consider using a less-than test. 1599729 9 385949 graouts 2019-12-18 03:00:01 -0800 Created attachment 385949 Patch 1599779 10 385949 commit-queue 2019-12-18 06:57:39 -0800 Comment on attachment 385949 Patch Clearing flags on attachment: 385949 Committed r253679: <https://trac.webkit.org/changeset/253679> 1599780 11 commit-queue 2019-12-18 06:57:40 -0800 All reviewed patches have been landed. Closing bug. 1601151 12 graouts 2019-12-22 03:29:11 -0800 Looking good so far, no failures since this change. 1601397 13 graouts 2019-12-24 04:39:23 -0800 No failure in 6 days, I think the flakiness is addressed. 385949 2019-12-18 03:00:01 -0800 2019-12-18 06:57:39 -0800 Patch bug-205299-20191218120000.patch text/plain 3518 graouts U3VidmVyc2lvbiBSZXZpc2lvbjogMjUzNjczCmRpZmYgLS1naXQgYS9MYXlvdXRUZXN0cy9DaGFu Z2VMb2cgYi9MYXlvdXRUZXN0cy9DaGFuZ2VMb2cKaW5kZXggYzQzMjQ5NWM5NjU5Zjg5NzY3YTk0 YjM5MjU3YjdlNmQ4NGNjZGJkMS4uNmUwZjM1ZWViYTY1NzIxMDBhMmI2NDc0NjM2ODVhZWEyMjEy OTk1YiAxMDA2NDQKLS0tIGEvTGF5b3V0VGVzdHMvQ2hhbmdlTG9nCisrKyBiL0xheW91dFRlc3Rz L0NoYW5nZUxvZwpAQCAtMSwzICsxLDIyIEBACisyMDE5LTEyLTE4ICBBbnRvaW5lIFF1aW50ICA8 Z3Jhb3V0c0BhcHBsZS5jb20+CisKKyAgICAgICAgUkVHUkVTU0lPTjogYW5pbWF0aW9ucy9sZWFr LWRvY3VtZW50LXdpdGgtY3NzLWFuaW1hdGlvbi5odG1sIGlzIHZlcnkgZmxha3kKKyAgICAgICAg aHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTIwNTI5OQorCisgICAgICAg IFdoaWxlIHRoaXMgdGVzdCBoYXMgcHJvdmVuIHRvIGJlIHNvbGlkIHJ1bm5pbmcgaW4gaXNvbGF0 aW9uLCBpdCBhcHBlYXJzIHRoYXQgdGhlIGRpZmZlcmVuY2Ugb2YgbGl2ZSBkb2N1bWVudHMKKyAg ICAgICAgYmVmb3JlIGFuZCBhZnRlciB0aGUgaWZyYW1lIGhhcyBiZWVuIGRlc3Ryb3llZCBtYXkg YmUgbW9yZSB0aGFuIG9uZS4gSW4gZmFjdCwgaW4gcmVwb3J0ZWQgZmFpbHVyZXMsIHRoZXJlIGFs d2F5cworICAgICAgICBzZWVtcyB0byBiZSAyIGxpdmUgZG9jdW1lbnRzIGZld2VyIGFmdGVyIHRo YW4gYmVmb3JlLCBpbnN0ZWFkIG9mIHRoZSBleHBlY3RlZCBzaW5nbGUgZG9jdW1lbnQuCisKKyAg ICAgICAgU2luY2UgaW50ZXJuYWxzLm51bWJlck9mTGl2ZURvY3VtZW50cygpIHJldHVybnMgdGhl IG51bWJlciBvZiBsaXZlIGRvY3VtZW50cyB0aHJvdWdob3V0IHRoZSBsaWZlIG9mIHRoZQorICAg ICAgICBXZWIgcHJvY2Vzcywgd2Ugbm93IGZvcmNlIGdhcmJhZ2UgY29sbGVjdGlvbiBldmVyeSB0 aW1lIHdlIHF1ZXJ5IGludGVybmFscy5udW1iZXJPZkxpdmVEb2N1bWVudHMoKS4gVGhpcworICAg ICAgICBzaG91bGQgZW5zdXJlIHRoYXQgZG9jdW1lbnRzIHBlbmRpbmcgZGVzdHJ1Y3Rpb24gZHVl IHRvIGdhcmJhZ2UgY29sbGVjdGlvbiB0byBiZSBkZXN0cm95ZWQgYmVmb3JlIHdlIHRha2UKKyAg ICAgICAgYSBzbmFwc2hvcnQgb2YgdGhlIG51bWJlciBvZiBsaXZlIGRvY3VtZW50cy4KKworICAg ICAgICBSZXZpZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICAqIFRlc3RFeHBlY3Rh dGlvbnM6CisgICAgICAgICogYW5pbWF0aW9ucy9sZWFrLWRvY3VtZW50LXdpdGgtY3NzLWFuaW1h dGlvbi5odG1sOgorCiAyMDE5LTEyLTE3ICBGdWppaSBIaXJvbm9yaSAgPEhpcm9ub3JpLkZ1amlp QHNvbnkuY29tPgogCiAgICAgICAgIFtpT1NdIEFyYWJpYyB0ZXh0IG9uIFdpa2lwZWRpYSBpcyBz aG93biBhcyBib3hlcwpkaWZmIC0tZ2l0IGEvTGF5b3V0VGVzdHMvVGVzdEV4cGVjdGF0aW9ucyBi L0xheW91dFRlc3RzL1Rlc3RFeHBlY3RhdGlvbnMKaW5kZXggMDNkODJkY2M5MDVhMWM0MzRmNzg5 MzJlMTgxNjIyNDc3MGNlOTRlOS4uMGZmYTc2ODJjNGVkNjY5ZGUxZmU1MTE4NGQ5ZTcxMGNjMzll YjViNiAxMDA2NDQKLS0tIGEvTGF5b3V0VGVzdHMvVGVzdEV4cGVjdGF0aW9ucworKysgYi9MYXlv dXRUZXN0cy9UZXN0RXhwZWN0YXRpb25zCkBAIC0zODE0LDUgKzM4MTQsMyBAQCB3ZWJnbC8xLjAu My9jb25mb3JtYW5jZS9yZW5kZXJidWZmZXJzL2ZlZWRiYWNrLWxvb3AuaHRtbCBbIFNraXAgXQog d2ViZ2wvMS4wLjMvY29uZm9ybWFuY2UvdGV4dHVyZXMvY29weS10ZXgtaW1hZ2UtMmQtZm9ybWF0 cy5odG1sIFsgU2tpcCBdCiB3ZWJnbC8xLjAuMy9jb25mb3JtYW5jZS9jYW52YXMvcmFwaWQtcmVz aXppbmcuaHRtbCBbIFNraXAgXQogd2ViZ2wvMS4wLjMvY29uZm9ybWFuY2UvZXh0ZW5zaW9ucy93 ZWJnbC1kcmF3LWJ1ZmZlcnMuaHRtbCBbIFNraXAgXQotCi13ZWJraXQub3JnL2IvMjA1Mjk5IGFu aW1hdGlvbnMvbGVhay1kb2N1bWVudC13aXRoLWNzcy1hbmltYXRpb24uaHRtbCBbIFBhc3MgRmFp bHVyZSBdCmRpZmYgLS1naXQgYS9MYXlvdXRUZXN0cy9hbmltYXRpb25zL2xlYWstZG9jdW1lbnQt d2l0aC1jc3MtYW5pbWF0aW9uLmh0bWwgYi9MYXlvdXRUZXN0cy9hbmltYXRpb25zL2xlYWstZG9j dW1lbnQtd2l0aC1jc3MtYW5pbWF0aW9uLmh0bWwKaW5kZXggZGI1YjNkMTQ4YzM1YjQwODY1MjZm YzFiNjlmNDVlMDE2MmU2YzA2Yi4uMDVmOGZmMTFiYmRmNWUyNmJiNzUxYTg3ZjBhMTdkOGFhZTVj NTFmMSAxMDA2NDQKLS0tIGEvTGF5b3V0VGVzdHMvYW5pbWF0aW9ucy9sZWFrLWRvY3VtZW50LXdp dGgtY3NzLWFuaW1hdGlvbi5odG1sCisrKyBiL0xheW91dFRlc3RzL2FuaW1hdGlvbnMvbGVhay1k b2N1bWVudC13aXRoLWNzcy1hbmltYXRpb24uaHRtbApAQCAtMjIsMTggKzIyLDIxIEBAIGZ1bmN0 aW9uIHJ1blRlc3QoKSB7CiAgICAgICAgIGlmIChmcmFtZS5zcmMgPT09ICdhYm91dDpibGFuaycp CiAgICAgICAgICAgICByZXR1cm4gdHJ1ZTsKIAotICAgICAgICBudW1iZXJPZkxpdmVEb2N1bWVu dHNBZnRlcklmcmFtZUxvYWRlZCA9IGludGVybmFscy5udW1iZXJPZkxpdmVEb2N1bWVudHMoKTsK LSAgICAgICAgZGVidWcoIlRoZSBpZnJhbWUgaGFzIGZpbmlzaGVkIGxvYWRpbmcuIik7Ci0KLSAg ICAgICAgZnJhbWUucmVtb3ZlKCk7Ci0gICAgICAgIGZyYW1lID0gbnVsbDsKLQogICAgICAgICBn YygpOwogICAgICAgICBzZXRUaW1lb3V0KGZ1bmN0aW9uICgpIHsKLSAgICAgICAgICAgIGRlYnVn KCJUaGUgaWZyYW1lIGhhcyBiZWVuIGRlc3Ryb3llZC4iKTsKLSAgICAgICAgICAgIHNob3VsZEJl KCJpbnRlcm5hbHMubnVtYmVyT2ZMaXZlRG9jdW1lbnRzKCkiLCAibnVtYmVyT2ZMaXZlRG9jdW1l bnRzQWZ0ZXJJZnJhbWVMb2FkZWQgLSAxIik7Ci0gICAgICAgICAgICBkZWJ1ZygiIik7Ci0gICAg ICAgICAgICBmaW5pc2hKU1Rlc3QoKTsKKyAgICAgICAgICAgIG51bWJlck9mTGl2ZURvY3VtZW50 c0FmdGVySWZyYW1lTG9hZGVkID0gaW50ZXJuYWxzLm51bWJlck9mTGl2ZURvY3VtZW50cygpOwor ICAgICAgICAgICAgZGVidWcoIlRoZSBpZnJhbWUgaGFzIGZpbmlzaGVkIGxvYWRpbmcuIik7CisK KyAgICAgICAgICAgIGZyYW1lLnJlbW92ZSgpOworICAgICAgICAgICAgZnJhbWUgPSBudWxsOwor CisgICAgICAgICAgICBnYygpOworICAgICAgICAgICAgc2V0VGltZW91dChmdW5jdGlvbiAoKSB7 CisgICAgICAgICAgICAgICAgZGVidWcoIlRoZSBpZnJhbWUgaGFzIGJlZW4gZGVzdHJveWVkLiIp OworICAgICAgICAgICAgICAgIHNob3VsZEJlKCJpbnRlcm5hbHMubnVtYmVyT2ZMaXZlRG9jdW1l bnRzKCkiLCAibnVtYmVyT2ZMaXZlRG9jdW1lbnRzQWZ0ZXJJZnJhbWVMb2FkZWQgLSAxIik7Cisg ICAgICAgICAgICAgICAgZGVidWcoIiIpOworICAgICAgICAgICAgICAgIGZpbmlzaEpTVGVzdCgp OworICAgICAgICAgICAgfSk7CiAgICAgICAgIH0pOwogICAgIH0KIAo=