233684 2021-11-30 23:04:34 -0800 [Performance test] Add test lots-of-self-painting-layers.html 2021-12-13 10:15:07 -0800 1 1 1 Unclassified WebKit Layout and Rendering WebKit Local Build Unspecified Unspecified RESOLVED FIXED InRadar P2 Normal --- 229802 1 cathiechen cathiechen bfulgham cdumez ews-watchlist jbedard rbuis rniwa simon.fraser webkit-bug-importer zalan zimmermann oldest_to_newest 1819104 0 cathiechen 2021-11-30 23:04:34 -0800 Add a test to track the repaint performance of self-painting layers. 1819111 1 445527 cathiechen 2021-11-30 23:09:50 -0800 Created attachment 445527 Patch 1819119 2 cathiechen 2021-11-30 23:30:54 -0800 The performance regression caused by self-painting layer. Self-paint layers(position: relative): :Time -> [515.0000000000018, 452, 452, 450, 453, 660, 657, 655.0000000000018, 656.0000000000018, 652, 673.9999999999982, 657.0000000000018, 661, 656, 661, 659, 660.9999999999964, 663.9999999999964, 659, 675] ms mean: 611.4499999999998 ms median: 657.0000000000009 ms stdev: 88.24038638556408 ms min: 450 ms max: 675 ms Normal flow layers(position: static): :Time -> [471, 342, 343, 335, 335, 402, 396, 383, 390, 385, 390, 387, 382, 381, 384, 379, 381.0000000000018, 385.0000000000018, 381, 381] ms mean: 380.6500000000002 ms median: 382.5 ms stdev: 29.207560304608954 ms min: 335 ms max: 471 ms 1821423 3 webkit-bug-importer 2021-12-07 23:05:22 -0800 <rdar://problem/86193972> 1822569 4 zimmermann 2021-12-10 01:14:35 -0800 I wouldn't call it a performance regression, after all it's a different mode. Creating N layers each rendering M elements vs. rendering one layer that renders N*M elements is naturally different: there's non-negligible amount of overhead associated with the management of layers, such as updateLayerPosition(s), etc. Or do you call it a regression, since it was better at some point? 1823446 5 cathiechen 2021-12-13 01:57:24 -0800 (In reply to Nikolas Zimmermann from comment #4) > I wouldn't call it a performance regression, after all it's a different > mode. Creating N layers each rendering M elements vs. rendering one layer > that renders N*M elements is naturally different: there's non-negligible > amount of overhead associated with the management of layers, such as > updateLayerPosition(s), etc. > Yeah, I agree with you, they are different modes. I called this as a regression, because we were addressing a performance regression for CSS contain. Like position: relative, CSS contain creates self-painting layers, the overhead of self-painting layers caused the regression. Because the regression is not specific to CSS contain, and CSS contain is not enabled at that time. So we created this test, and keep calling it as a regression. I guess we can call this test as a tracking for the overhead cost of self-painting layers. What do you think? > Or do you call it a regression, since it was better at some point? Nope. 1823468 6 zimmermann 2021-12-13 04:14:20 -0800 (In reply to cathiechen from comment #5) > (In reply to Nikolas Zimmermann from comment #4) > > I wouldn't call it a performance regression, after all it's a different > > mode. Creating N layers each rendering M elements vs. rendering one layer > > that renders N*M elements is naturally different: there's non-negligible > > amount of overhead associated with the management of layers, such as > > updateLayerPosition(s), etc. > > > > Yeah, I agree with you, they are different modes. > > I called this as a regression, because we were addressing a performance > regression for CSS contain. Like position: relative, CSS contain creates > self-painting layers, the overhead of self-painting layers caused the > regression. > Because the regression is not specific to CSS contain, and CSS contain is > not enabled at that time. So we created this test, and keep calling it as a > regression. > > I guess we can call this test as a tracking for the overhead cost of > self-painting layers. What do you think? Agreed, I just found the wording regression confusing and thought you were really fixing a regression from a certain revision, that's not the case, so everything is fine. Thanks, for looking at layer performance, very important for LBSE to migitate the self-painting layer overheads as much as possible. 1823545 7 447008 cathiechen 2021-12-13 08:47:28 -0800 Created attachment 447008 Patch 1823553 8 cathiechen 2021-12-13 09:00:06 -0800 (In reply to Nikolas Zimmermann from comment #6) > (In reply to cathiechen from comment #5) > > (In reply to Nikolas Zimmermann from comment #4) > > > I wouldn't call it a performance regression, after all it's a different > > > mode. Creating N layers each rendering M elements vs. rendering one layer > > > that renders N*M elements is naturally different: there's non-negligible > > > amount of overhead associated with the management of layers, such as > > > updateLayerPosition(s), etc. > > > > > > > Yeah, I agree with you, they are different modes. > > > > I called this as a regression, because we were addressing a performance > > regression for CSS contain. Like position: relative, CSS contain creates > > self-painting layers, the overhead of self-painting layers caused the > > regression. > > Because the regression is not specific to CSS contain, and CSS contain is > > not enabled at that time. So we created this test, and keep calling it as a > > regression. > > > > I guess we can call this test as a tracking for the overhead cost of > > self-painting layers. What do you think? > Agreed, I just found the wording regression confusing and thought you were > really fixing a regression from a certain revision, that's not the case, so > everything is fine. > > Thanks, for looking at layer performance, very important for LBSE to > migitate the self-painting layer overheads as much as possible. Great! Thanks for pointing this out:) 1823592 9 ews-feeder 2021-12-13 10:15:02 -0800 Committed r286954 (245179@main): <https://commits.webkit.org/245179@main> All reviewed patches have been landed. Closing bug and clearing flags on attachment 447008. 445527 2021-11-30 23:09:50 -0800 2021-12-13 08:47:22 -0800 Patch bug-233684-20211201150947.patch text/plain 3738 cathiechen U3VidmVyc2lvbiBSZXZpc2lvbjogMjg2MTA4CmRpZmYgLS1naXQgYS9QZXJmb3JtYW5jZVRlc3Rz L0NoYW5nZUxvZyBiL1BlcmZvcm1hbmNlVGVzdHMvQ2hhbmdlTG9nCmluZGV4IDg5MjM5MmQ4N2Y3 NDAwZWJmYTk4ZjVhODY1M2M5ZmQ0ZTI3NTVmNTkuLjQwZDE4YzU5ZGQ1MGZlN2E0NzIwZDIzZTZj MDc2YTVjM2E2YzhkZjYgMTAwNjQ0Ci0tLSBhL1BlcmZvcm1hbmNlVGVzdHMvQ2hhbmdlTG9nCisr KyBiL1BlcmZvcm1hbmNlVGVzdHMvQ2hhbmdlTG9nCkBAIC0xLDMgKzEsMTggQEAKKzIwMjEtMTEt MzAgIENhdGhpZSBDaGVuICA8Y2F0aGllY2hlbkBpZ2FsaWEuY29tPgorCisgICAgICAgIFtQZXJm b3JtYW5jZSB0ZXN0XSBBZGQgdGVzdCBsb3RzLW9mLXNlbGYtcGFpbnRpbmctbGF5ZXJzLmh0bWwK KyAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTIzMzY4NAor CisgICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIFRoaXMgdGVz dCBpcyBhIHRyYW5zZm9ybSBvZiB0aGUgdGVzdCBbMV0gaW4gYmxpbmsuIEl0IHVzZXMgcG9zaXRp b246IHJlbGF0aXZlIHRvIGNyZWF0ZQorICAgICAgICBsb3RzIG9mIHNlbGYtcGFpbnRpbmcgbGF5 ZXJzLiBCeSBjaGFuZ2luZyB0aGUgZmlyc3QgbGF5ZXIncyBoZWlnaHQsIGl0IHRyYWNrcyB0aGUg cmVwYWludAorICAgICAgICBwZXJmb3JtYW5jZSBvZiB0aGVzZSBzZWxmLXBhaW50aW5nIGxheWVy cy4KKworICAgICAgICBbMV0gdGhpcmRfcGFydHkvYmxpbmsvcGVyZl90ZXN0cy9sYXlvdXQvY29u dGFpbi1jb250ZW50LXN0eWxlLWNoYW5nZS5odG1sCisKKyAgICAgICAgKiBQYWludC9sb3RzLW9m LXNlbGYtcGFpbnRpbmctbGF5ZXJzLmh0bWw6IEFkZGVkLgorCiAyMDIxLTExLTIxICBDYXRoaWUg Q2hlbiAgPGNhdGhpZWNoZW5AaWdhbGlhLmNvbT4KIAogICAgICAgICBbUGVyZm9ybWFuY2UgdGVz dF1bY3NzLWNvbnRhaW5dIEFkZCBjYXNlIHRvIHRlc3QgY29udGFpbjogbGF5b3V0IGluc2lkZSBh IGNvbXBsZXggZG9jdW1lbnQKZGlmZiAtLWdpdCBhL1BlcmZvcm1hbmNlVGVzdHMvUGFpbnQvbG90 cy1vZi1zZWxmLXBhaW50aW5nLWxheWVycy5odG1sIGIvUGVyZm9ybWFuY2VUZXN0cy9QYWludC9s b3RzLW9mLXNlbGYtcGFpbnRpbmctbGF5ZXJzLmh0bWwKbmV3IGZpbGUgbW9kZSAxMDA2NDQKaW5k ZXggMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMC4uMWE3YTBiNTQ5YTc1 YzU3ZThlNjA1ZmU4ODk1ZmY4MmI1NjQ4ZTRlZQotLS0gL2Rldi9udWxsCisrKyBiL1BlcmZvcm1h bmNlVGVzdHMvUGFpbnQvbG90cy1vZi1zZWxmLXBhaW50aW5nLWxheWVycy5odG1sCkBAIC0wLDAg KzEsNjcgQEAKKzwhRE9DVFlQRSBodG1sPgorPHNjcmlwdCBzcmM9Ii4uL3Jlc291cmNlcy9ydW5u ZXIuanMiPjwvc2NyaXB0PgorPHRpdGxlPlRyYWNraW5nIHRoZSBwZXJmb3JtYW5jZSBvZiBzZWxm LXBhaW50aW5nIGxheWVyczwvdGl0bGU+Cis8c3R5bGU+CisgICAgI2xpc3RDb250YWluZXIgewor ICAgICAgICBtYXJnaW46IDAgYXV0bzsKKyAgICAgICAgd2lkdGg6IDYwMHB4OworICAgICAgICBw b3NpdGlvbjogcmVsYXRpdmU7CisgICAgfQorCisgICAgLmxpc3RJdGVtIHsKKyAgICAgICAgb3V0 bGluZTogMnB4IHNvbGlkIGdyZWVuOworICAgICAgICBvdmVyZmxvdzogaGlkZGVuOworICAgICAg ICBwb3NpdGlvbjogcmVsYXRpdmU7CisgICAgfQorPC9zdHlsZT4KKzwvaGVhZD4KKworPGJvZHk+ CisgICAgPHByZSBpZD0ibG9nIj48L3ByZT4KKworICAgIDxkaXYgaWQ9Imxpc3RDb250YWluZXIi PjwvZGl2PgorICAgIDxzY3JpcHQ+CisgICAgICAgIGNvbnN0IExPUkVNX0lQU1VNID0gIkxvcmVt IGlwc3VtIGRvbG9yIHNpdCBhbWV0LCBjb25zZWN0ZXR1ciBhZGlwaXNjaW5nIGVsaXQsIHNlZCBk byBlaXVzbW9kIHRlbXBvciBpbmNpZGlkdW50IHV0IGxhYm9yZSBldCBkb2xvcmUgbWFnbmEgYWxp cXVhLiBVdCBlbmltIGFkIG1pbmltIHZlbmlhbSwgcXVpcyBub3N0cnVkIGV4ZXJjaXRhdGlvbiB1 bGxhbWNvIGxhYm9yaXMgbmlzaSB1dCBhbGlxdWlwIGV4IGVhIGNvbW1vZG8gY29uc2VxdWF0LiBE dWlzIGF1dGUgaXJ1cmUgZG9sb3IgaW4gcmVwcmVoZW5kZXJpdCBpbiB2b2x1cHRhdGUgdmVsaXQg ZXNzZSBjaWxsdW0gZG9sb3JlIGV1IGZ1Z2lhdCBudWxsYSBwYXJpYXR1ci4gRXhjZXB0ZXVyIHNp bnQgb2NjYWVjYXQgY3VwaWRhdGF0IG5vbiBwcm9pZGVudCwgc3VudCBpbiBjdWxwYSBxdWkgb2Zm aWNpYSBkZXNlcnVudCBtb2xsaXQgYW5pbSBpZCBlc3QgbGFib3J1bS4iOworCisgICAgICAgIGxl dCBsaXN0Q29udGFpbmVyID0gbnVsbDsKKyAgICAgICAgZnVuY3Rpb24gY3JlYXRlTGlzdEl0ZW0o KSB7CisgICAgICAgICAgICBsZXQgaXRlbSA9IGRvY3VtZW50LmNyZWF0ZUVsZW1lbnQoImRpdiIp OworICAgICAgICAgICAgaXRlbS5jbGFzc0xpc3QuYWRkKCJsaXN0SXRlbSIpOworICAgICAgICAg ICAgaXRlbS50ZXh0Q29udGVudCA9IExPUkVNX0lQU1VNLnN1YnN0cigKKyAgICAgICAgICAgICAg ICBNYXRoLmZsb29yKE1hdGgucmFuZG9tKCkgKiAxMDApLCBNYXRoLmZsb29yKE1hdGgucmFuZG9t KCkgKiAyMDApICsgMTUwKTsKKyAgICAgICAgICAgIGxpc3RDb250YWluZXIuYXBwZW5kQ2hpbGQo aXRlbSk7CisgICAgICAgIH0KKworICAgICAgICBmdW5jdGlvbiBzZXR1cFRlc3QoKSB7CisgICAg ICAgICAgICBsaXN0Q29udGFpbmVyID0gZG9jdW1lbnQuZ2V0RWxlbWVudEJ5SWQoImxpc3RDb250 YWluZXIiKTsKKyAgICAgICAgICAgIGZvciAobGV0IGkgPSAwOyBpIDwgMTAwMDA7ICsraSkKKyAg ICAgICAgICAgICAgICBjcmVhdGVMaXN0SXRlbSgpOworICAgICAgICAgICAgZG9jdW1lbnQuYm9k eS5nZXRCb3VuZGluZ0NsaWVudFJlY3QoKTsKKyAgICAgICAgICAgIHdpbmRvdy5yZXF1ZXN0QW5p bWF0aW9uRnJhbWUocnVuVGVzdCk7CisgICAgICAgIH0KKworICAgICAgICB2YXIgaXNEb25lID0g ZmFsc2U7CisgICAgICAgIFBlcmZUZXN0UnVubmVyLnByZXBhcmVUb01lYXN1cmVWYWx1ZXNBc3lu Yyh7IGRvbmU6IGRvbmUsIHVuaXQ6ICdtcycgfSk7CisgICAgICAgIGZ1bmN0aW9uIGRvbmUoKSB7 CisgICAgICAgICAgICBpc0RvbmUgPSB0cnVlOworICAgICAgICAgICAgbGlzdENvbnRhaW5lci5p bm5lckhUTUwgPSAiIjsKKyAgICAgICAgfQorCisgICAgICAgIHZhciBzdGFydFRpbWU7CisgICAg ICAgIHZhciBoZWlnaHQgPSAwOworICAgICAgICBmdW5jdGlvbiBydW5UZXN0KCkgeworICAgICAg ICAgICAgaWYgKHN0YXJ0VGltZSkKKyAgICAgICAgICAgICAgICBQZXJmVGVzdFJ1bm5lci5tZWFz dXJlVmFsdWVBc3luYyhQZXJmVGVzdFJ1bm5lci5ub3coKSAtIHN0YXJ0VGltZSk7CisKKyAgICAg ICAgICAgIGlmICghaXNEb25lKSB7CisgICAgICAgICAgICAgICAgc3RhcnRUaW1lID0gUGVyZlRl c3RSdW5uZXIubm93KCk7CisgICAgICAgICAgICAgICAgbGlzdENvbnRhaW5lci5maXJzdENoaWxk LnN0eWxlLmhlaWdodCA9IGhlaWdodCArICJweCI7CisgICAgICAgICAgICAgICAgaGVpZ2h0Kys7 CisgICAgICAgICAgICAgICAgd2luZG93LnJlcXVlc3RBbmltYXRpb25GcmFtZShydW5UZXN0KTsK KyAgICAgICAgICAgIH0KKyAgICAgICAgfQorCisgICAgICAgIG9ubG9hZCA9IHNldHVwVGVzdDsK KyAgICA8L3NjcmlwdD4KKzwvYm9keT4KKzwvaHRtbD4K 447008 2021-12-13 08:47:28 -0800 2021-12-13 10:15:05 -0800 Patch bug-233684-20211214004727.patch text/plain 3705 cathiechen U3VidmVyc2lvbiBSZXZpc2lvbjogMjg2OTQ4CmRpZmYgLS1naXQgYS9QZXJmb3JtYW5jZVRlc3Rz L0NoYW5nZUxvZyBiL1BlcmZvcm1hbmNlVGVzdHMvQ2hhbmdlTG9nCmluZGV4IGYwMWFiOTBiNDc3 ZDZmZGU5NTcwNDEwYTRmZGVkMGEwMjNlNmJhODUuLjA5M2ZjZmY4ZmY0ZTEwMDc2N2I5NmViNWY1 MGQzMDBhODVmZWY3YTAgMTAwNjQ0Ci0tLSBhL1BlcmZvcm1hbmNlVGVzdHMvQ2hhbmdlTG9nCisr KyBiL1BlcmZvcm1hbmNlVGVzdHMvQ2hhbmdlTG9nCkBAIC0xLDMgKzEsMTggQEAKKzIwMjEtMTIt MTMgIENhdGhpZSBDaGVuICA8Y2F0aGllY2hlbkBpZ2FsaWEuY29tPgorCisgICAgICAgIFtQZXJm b3JtYW5jZSB0ZXN0XSBBZGQgdGVzdCBsb3RzLW9mLXNlbGYtcGFpbnRpbmctbGF5ZXJzLmh0bWwK KyAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTIzMzY4NAor CisgICAgICAgIFJldmlld2VkIGJ5IFJvYiBCdWlzLgorCisgICAgICAgIFRoaXMgdGVzdCBpcyBh IHRyYW5zZm9ybSBvZiB0aGUgdGVzdCBbMV0gaW4gYmxpbmsuIEl0IHVzZXMgcG9zaXRpb246IHJl bGF0aXZlIHRvIGNyZWF0ZQorICAgICAgICBsb3RzIG9mIHNlbGYtcGFpbnRpbmcgbGF5ZXJzLiBC eSBjaGFuZ2luZyB0aGUgZmlyc3QgbGF5ZXIncyBoZWlnaHQsIGl0IHRyYWNrcyB0aGUgcmVwYWlu dAorICAgICAgICBwZXJmb3JtYW5jZSBvZiB0aGVzZSBzZWxmLXBhaW50aW5nIGxheWVycy4KKwor ICAgICAgICBbMV0gdGhpcmRfcGFydHkvYmxpbmsvcGVyZl90ZXN0cy9sYXlvdXQvY29udGFpbi1j b250ZW50LXN0eWxlLWNoYW5nZS5odG1sCisKKyAgICAgICAgKiBQYWludC9sb3RzLW9mLXNlbGYt cGFpbnRpbmctbGF5ZXJzLmh0bWw6IEFkZGVkLgorCiAyMDIxLTEyLTA4ICBBbGFuIEJ1anRhcyAg PHphbGFuQGFwcGxlLmNvbT4KIAogICAgICAgICBBZGQgbGF5b3V0IHBlcmZvcm1hbmNlIHRlc3Qg Zm9yIHNpbXBsZSBiaWRpIGNvbnRlbnQgd2l0aCBpbmxpbmUgYm94ZXMKZGlmZiAtLWdpdCBhL1Bl cmZvcm1hbmNlVGVzdHMvUGFpbnQvbG90cy1vZi1zZWxmLXBhaW50aW5nLWxheWVycy5odG1sIGIv UGVyZm9ybWFuY2VUZXN0cy9QYWludC9sb3RzLW9mLXNlbGYtcGFpbnRpbmctbGF5ZXJzLmh0bWwK bmV3IGZpbGUgbW9kZSAxMDA2NDQKaW5kZXggMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAw MDAwMDAwMDAwMC4uMWE3YTBiNTQ5YTc1YzU3ZThlNjA1ZmU4ODk1ZmY4MmI1NjQ4ZTRlZQotLS0g L2Rldi9udWxsCisrKyBiL1BlcmZvcm1hbmNlVGVzdHMvUGFpbnQvbG90cy1vZi1zZWxmLXBhaW50 aW5nLWxheWVycy5odG1sCkBAIC0wLDAgKzEsNjcgQEAKKzwhRE9DVFlQRSBodG1sPgorPHNjcmlw dCBzcmM9Ii4uL3Jlc291cmNlcy9ydW5uZXIuanMiPjwvc2NyaXB0PgorPHRpdGxlPlRyYWNraW5n IHRoZSBwZXJmb3JtYW5jZSBvZiBzZWxmLXBhaW50aW5nIGxheWVyczwvdGl0bGU+Cis8c3R5bGU+ CisgICAgI2xpc3RDb250YWluZXIgeworICAgICAgICBtYXJnaW46IDAgYXV0bzsKKyAgICAgICAg d2lkdGg6IDYwMHB4OworICAgICAgICBwb3NpdGlvbjogcmVsYXRpdmU7CisgICAgfQorCisgICAg Lmxpc3RJdGVtIHsKKyAgICAgICAgb3V0bGluZTogMnB4IHNvbGlkIGdyZWVuOworICAgICAgICBv dmVyZmxvdzogaGlkZGVuOworICAgICAgICBwb3NpdGlvbjogcmVsYXRpdmU7CisgICAgfQorPC9z dHlsZT4KKzwvaGVhZD4KKworPGJvZHk+CisgICAgPHByZSBpZD0ibG9nIj48L3ByZT4KKworICAg IDxkaXYgaWQ9Imxpc3RDb250YWluZXIiPjwvZGl2PgorICAgIDxzY3JpcHQ+CisgICAgICAgIGNv bnN0IExPUkVNX0lQU1VNID0gIkxvcmVtIGlwc3VtIGRvbG9yIHNpdCBhbWV0LCBjb25zZWN0ZXR1 ciBhZGlwaXNjaW5nIGVsaXQsIHNlZCBkbyBlaXVzbW9kIHRlbXBvciBpbmNpZGlkdW50IHV0IGxh Ym9yZSBldCBkb2xvcmUgbWFnbmEgYWxpcXVhLiBVdCBlbmltIGFkIG1pbmltIHZlbmlhbSwgcXVp cyBub3N0cnVkIGV4ZXJjaXRhdGlvbiB1bGxhbWNvIGxhYm9yaXMgbmlzaSB1dCBhbGlxdWlwIGV4 IGVhIGNvbW1vZG8gY29uc2VxdWF0LiBEdWlzIGF1dGUgaXJ1cmUgZG9sb3IgaW4gcmVwcmVoZW5k ZXJpdCBpbiB2b2x1cHRhdGUgdmVsaXQgZXNzZSBjaWxsdW0gZG9sb3JlIGV1IGZ1Z2lhdCBudWxs YSBwYXJpYXR1ci4gRXhjZXB0ZXVyIHNpbnQgb2NjYWVjYXQgY3VwaWRhdGF0IG5vbiBwcm9pZGVu dCwgc3VudCBpbiBjdWxwYSBxdWkgb2ZmaWNpYSBkZXNlcnVudCBtb2xsaXQgYW5pbSBpZCBlc3Qg bGFib3J1bS4iOworCisgICAgICAgIGxldCBsaXN0Q29udGFpbmVyID0gbnVsbDsKKyAgICAgICAg ZnVuY3Rpb24gY3JlYXRlTGlzdEl0ZW0oKSB7CisgICAgICAgICAgICBsZXQgaXRlbSA9IGRvY3Vt ZW50LmNyZWF0ZUVsZW1lbnQoImRpdiIpOworICAgICAgICAgICAgaXRlbS5jbGFzc0xpc3QuYWRk KCJsaXN0SXRlbSIpOworICAgICAgICAgICAgaXRlbS50ZXh0Q29udGVudCA9IExPUkVNX0lQU1VN LnN1YnN0cigKKyAgICAgICAgICAgICAgICBNYXRoLmZsb29yKE1hdGgucmFuZG9tKCkgKiAxMDAp LCBNYXRoLmZsb29yKE1hdGgucmFuZG9tKCkgKiAyMDApICsgMTUwKTsKKyAgICAgICAgICAgIGxp c3RDb250YWluZXIuYXBwZW5kQ2hpbGQoaXRlbSk7CisgICAgICAgIH0KKworICAgICAgICBmdW5j dGlvbiBzZXR1cFRlc3QoKSB7CisgICAgICAgICAgICBsaXN0Q29udGFpbmVyID0gZG9jdW1lbnQu Z2V0RWxlbWVudEJ5SWQoImxpc3RDb250YWluZXIiKTsKKyAgICAgICAgICAgIGZvciAobGV0IGkg PSAwOyBpIDwgMTAwMDA7ICsraSkKKyAgICAgICAgICAgICAgICBjcmVhdGVMaXN0SXRlbSgpOwor ICAgICAgICAgICAgZG9jdW1lbnQuYm9keS5nZXRCb3VuZGluZ0NsaWVudFJlY3QoKTsKKyAgICAg ICAgICAgIHdpbmRvdy5yZXF1ZXN0QW5pbWF0aW9uRnJhbWUocnVuVGVzdCk7CisgICAgICAgIH0K KworICAgICAgICB2YXIgaXNEb25lID0gZmFsc2U7CisgICAgICAgIFBlcmZUZXN0UnVubmVyLnBy ZXBhcmVUb01lYXN1cmVWYWx1ZXNBc3luYyh7IGRvbmU6IGRvbmUsIHVuaXQ6ICdtcycgfSk7Cisg ICAgICAgIGZ1bmN0aW9uIGRvbmUoKSB7CisgICAgICAgICAgICBpc0RvbmUgPSB0cnVlOworICAg ICAgICAgICAgbGlzdENvbnRhaW5lci5pbm5lckhUTUwgPSAiIjsKKyAgICAgICAgfQorCisgICAg ICAgIHZhciBzdGFydFRpbWU7CisgICAgICAgIHZhciBoZWlnaHQgPSAwOworICAgICAgICBmdW5j dGlvbiBydW5UZXN0KCkgeworICAgICAgICAgICAgaWYgKHN0YXJ0VGltZSkKKyAgICAgICAgICAg ICAgICBQZXJmVGVzdFJ1bm5lci5tZWFzdXJlVmFsdWVBc3luYyhQZXJmVGVzdFJ1bm5lci5ub3co KSAtIHN0YXJ0VGltZSk7CisKKyAgICAgICAgICAgIGlmICghaXNEb25lKSB7CisgICAgICAgICAg ICAgICAgc3RhcnRUaW1lID0gUGVyZlRlc3RSdW5uZXIubm93KCk7CisgICAgICAgICAgICAgICAg bGlzdENvbnRhaW5lci5maXJzdENoaWxkLnN0eWxlLmhlaWdodCA9IGhlaWdodCArICJweCI7Cisg ICAgICAgICAgICAgICAgaGVpZ2h0Kys7CisgICAgICAgICAgICAgICAgd2luZG93LnJlcXVlc3RB bmltYXRpb25GcmFtZShydW5UZXN0KTsKKyAgICAgICAgICAgIH0KKyAgICAgICAgfQorCisgICAg ICAgIG9ubG9hZCA9IHNldHVwVGVzdDsKKyAgICA8L3NjcmlwdD4KKzwvYm9keT4KKzwvaHRtbD4K