184062 2018-03-27 17:14:41 -0700 MSVC __forceinline slows down JSC release build fivefold after r229391 2018-03-29 19:47:40 -0700 1 1 1 Unclassified WebKit Web Template Framework WebKit Nightly Build Unspecified Unspecified RESOLVED FIXED https://bugs.webkit.org/show_bug.cgi?id=184166 InRadar P2 Normal --- 184128 1 ross.kirsling ross.kirsling achristensen basuke benjamin bfulgham cdumez chris.reid cmarcelo commit-queue dbates don.olmstead ews-watchlist jfbastien keith_miller mark.lam msaboff pvollan rniwa saam webkit-bug-importer ysuzuki oldest_to_newest 1409733 0 ross.kirsling 2018-03-27 17:14:41 -0700 After the TMP refactor in r229391 (https://bugs.webkit.org/show_bug.cgi?id=183263), the WinCairo Release build takes 50 minutes to compile JSC alone, compared to only 10 minutes prior to the patch. This is an MSVC-specific issue -- their optimizer has trouble dealing with "too liberal a use of __forceinline". (https://blogs.msdn.microsoft.com/vcblog/2018/01/04/visual-studio-2017-throughput-improvements-and-advice/) We have reported the issue to Microsoft, but in the meantime it would seem wise to define ALWAYS_INLINE as simply "inline" on Windows, which brings the compile time back down to 10 minutes. 1409735 1 336631 ross.kirsling 2018-03-27 17:17:34 -0700 Created attachment 336631 Patch 1409777 2 jfbastien 2018-03-27 21:32:05 -0700 That seems like a pretty wide net to cast to avoid the force-inline problem with the TMP code. Could you instead have a TMP_ALWAYS_INLINE or something like that which only applies to that code? On the other hand, I don't know how much tuning has been done for the Windows build and inlining, so the heuristics we've applied may be totally wrong. Do we have developers who measure whether this changes pessimizes performance? 1409879 3 ross.kirsling 2018-03-28 09:45:24 -0700 (In reply to JF Bastien from comment #2) > That seems like a pretty wide net to cast to avoid the force-inline problem > with the TMP code. Could you instead have a TMP_ALWAYS_INLINE or something > like that which only applies to that code? That might be a possibility. I guess there's a similar thing going on here: https://github.com/WebKit/webkit/blob/master/Source/WebCore/platform/graphics/FormatConverter.cpp#L39-L43 (...actually I wonder if this one's still necessary in VS2017, hehe.) > On the other hand, I don't know how much tuning has been done for the > Windows build and inlining, so the heuristics we've applied may be totally > wrong. Do we have developers who measure whether this changes pessimizes > performance? That would be great to know, especially as this should apply equally to AppleWin. I could look into performance test comparisons over here as well. 1409915 4 jfbastien 2018-03-28 10:47:44 -0700 (In reply to Ross Kirsling from comment #3) > (In reply to JF Bastien from comment #2) > > That seems like a pretty wide net to cast to avoid the force-inline problem > > with the TMP code. Could you instead have a TMP_ALWAYS_INLINE or something > > like that which only applies to that code? > > That might be a possibility. I guess there's a similar thing going on here: > https://github.com/WebKit/webkit/blob/master/Source/WebCore/platform/ > graphics/FormatConverter.cpp#L39-L43 > (...actually I wonder if this one's still necessary in VS2017, hehe.) > > > On the other hand, I don't know how much tuning has been done for the > > Windows build and inlining, so the heuristics we've applied may be totally > > wrong. Do we have developers who measure whether this changes pessimizes > > performance? > > That would be great to know, especially as this should apply equally to > AppleWin. I could look into performance test comparisons over here as well. Alex / Per / Brent are more informed about WebKit on Windows than I am, so I'll defer to them. 1409918 5 336631 achristensen 2018-03-28 10:52:29 -0700 Comment on attachment 336631 Patch I agree with JF that this is too drastic a change. I think we should instead do something like I needed to do in https://bugs.webkit.org/show_bug.cgi?id=125808 and only remove the __forceinline on only the problematic inlines. There are a lot of instances of ALWAYS_INLINE that we probably want to keep. 1409925 6 ross.kirsling 2018-03-28 11:01:08 -0700 (In reply to Alex Christensen from comment #5) > Comment on attachment 336631 [details] > Patch > > I agree with JF that this is too drastic a change. I think we should > instead do something like I needed to do in > https://bugs.webkit.org/show_bug.cgi?id=125808 and only remove the > __forceinline on only the problematic inlines. There are a lot of instances > of ALWAYS_INLINE that we probably want to keep. Sounds good. Just thought I'd kick off the conversation from the naive approach. :) 1410043 7 ross.kirsling 2018-03-28 14:01:35 -0700 It appears that the issue is indeed quite localized and must have to do with MSVC getting overwhelmed by recursive templates in particular. I've confirmed that: 1. Using non-forced inline for every `setupArgumentsImpl` overload results in a 9.5-minute build. https://github.com/WebKit/webkit/blob/master/Source/JavaScriptCore/jit/CCallHelpers.h#L318-L464 2. Using non-forced inline for `marshallArgumentRegister` alone results in an 11-minute build. https://github.com/WebKit/webkit/blob/master/Source/JavaScriptCore/jit/CCallHelpers.h#L299 These are both potentially acceptable -- does anyone have a clear preference? Perhaps a recommendation of which test(s) would be most useful for perf impact? 1410071 8 achristensen 2018-03-28 15:09:44 -0700 The least invasive change would be to make marshallArgumentRegister non-forced-inline on MSVC only. 1410082 9 ross.kirsling 2018-03-28 15:31:59 -0700 (In reply to Alex Christensen from comment #8) > The least invasive change would be to make marshallArgumentRegister > non-forced-inline on MSVC only. Sounds good. FWIW, I also checked Sunspider benchmarks, which (unsurprisingly) support the same conclusion: Collected 4 samples per benchmark/VM, with 4 VM invocations per benchmark. Emitted a call to gc() between sample measurements. Used 1 benchmark iteration per VM invocation for warm-up. Used the jsc-specific preciseTime() function to get microsecond-level timing. Reporting benchmark execution times with 95% confidence intervals in milliseconds. marshallArgumentRegister setupArgumentsImpl 3d-cube 17.1135+-1.2752 ? 17.4554+-1.8218 ? might be 1.0200x slower 3d-morph 12.1973+-0.8931 ? 13.6648+-2.0364 ? might be 1.1203x slower 3d-raytrace 23.5151+-2.4159 ? 25.1138+-7.1721 ? might be 1.0680x slower access-binary-trees 8.3303+-0.5779 ? 8.7787+-2.4330 ? might be 1.0538x slower access-fannkuch 14.8218+-2.2912 ? 15.4262+-7.4340 ? might be 1.0408x slower access-nbody 11.0309+-1.1147 10.1383+-0.2915 might be 1.0880x faster access-nsieve 6.6365+-0.8363 6.3742+-0.4838 might be 1.0411x faster bitops-3bit-bits-in-byte 5.1899+-0.1658 ? 5.8102+-1.2525 ? might be 1.1195x slower bitops-bits-in-byte 7.3887+-2.3269 6.6924+-0.1452 might be 1.1040x faster bitops-bitwise-and 9.1475+-1.1505 ? 9.3076+-1.3356 ? might be 1.0175x slower bitops-nsieve-bits 7.7549+-1.3521 7.7289+-0.6407 controlflow-recursive 7.2060+-0.4298 ! 8.4953+-0.8057 ! definitely 1.1789x slower crypto-aes 19.2986+-8.2190 18.0129+-3.0947 might be 1.0714x faster crypto-md5 8.2644+-1.2072 7.7612+-1.1438 might be 1.0648x faster crypto-sha1 8.1423+-1.5990 7.9152+-0.6251 might be 1.0287x faster date-format-tofte 35.1650+-0.6483 ? 36.1490+-2.2741 ? might be 1.0280x slower date-format-xparb 21.6351+-2.0519 21.2189+-0.1581 might be 1.0196x faster math-cordic 11.2579+-2.4418 9.3053+-0.4511 might be 1.2098x faster math-partial-sums 13.2822+-1.8823 13.0983+-2.1215 might be 1.0140x faster math-spectral-norm 7.7226+-0.4115 7.6494+-0.2857 regexp-dna 11.4542+-0.2199 ? 12.5857+-3.0776 ? might be 1.0988x slower string-base64 13.4223+-1.0573 ? 15.1614+-2.0637 ? might be 1.1296x slower string-fasta 20.6008+-2.9576 ? 21.6987+-6.0305 ? might be 1.0533x slower string-tagcloud 32.4288+-0.5497 ? 34.3539+-3.1244 ? might be 1.0594x slower string-unpack-code 84.0179+-8.3223 ? 90.4980+-27.9551 ? might be 1.0771x slower string-validate-input 23.8311+-1.8061 ? 24.0583+-1.7425 ? <arithmetic> 16.9560+-0.5638 ? 17.4789+-0.6725 ? might be 1.0308x slower 1410109 10 336731 ross.kirsling 2018-03-28 16:20:42 -0700 Created attachment 336731 Patch 1410117 11 ross.kirsling 2018-03-28 16:48:42 -0700 Additionally, it looks like that "C1063 Fatal Error" no longer occurs for FormatConverter.cpp, so we should be able to restore the ALWAYS_INLINE there in a follow-up patch. 1410129 12 achristensen 2018-03-28 17:10:25 -0700 (In reply to Ross Kirsling from comment #11) > Additionally, it looks like that "C1063 Fatal Error" no longer occurs for > FormatConverter.cpp, so we should be able to restore the ALWAYS_INLINE there > in a follow-up patch. I've been meaning to look into whether we should just remove the ALWAYS_INLINE there. It could make the binary significantly smaller while de-optiizing the outer loop, which would probably be a performance gain. 1410140 13 336731 commit-queue 2018-03-28 17:48:01 -0700 Comment on attachment 336731 Patch Clearing flags on attachment: 336731 Committed r230062: <https://trac.webkit.org/changeset/230062> 1410141 14 commit-queue 2018-03-28 17:48:03 -0700 All reviewed patches have been landed. Closing bug. 1410142 15 webkit-bug-importer 2018-03-28 17:49:24 -0700 <rdar://problem/38980095> 1410207 16 rniwa 2018-03-29 00:33:08 -0700 Somehow this patch caused Mac port to break. We can't load apple.com now :( 1410209 17 commit-queue 2018-03-29 00:33:14 -0700 Re-opened since this is blocked by bug 184128 1410306 18 achristensen 2018-03-29 10:39:45 -0700 (In reply to Ryosuke Niwa from comment #16) > Somehow this patch caused Mac port to break. We can't load apple.com now :( This change makes no change on Mac. You have almost certainly misdiagnosed 1410386 19 rniwa 2018-03-29 13:48:55 -0700 (In reply to Alex Christensen from comment #18) > (In reply to Ryosuke Niwa from comment #16) > > Somehow this patch caused Mac port to break. We can't load apple.com now :( > > This change makes no change on Mac. You have almost certainly misdiagnosed You say that but Safari builds before this patch loads web pages file, and Safari builds after this patch cause web content process to instantly crash. The crash reproduces at r230069 but doesn't after the rollout in r230070. So your assertion that this change makes no change on Mac is obviously wrong. It certainly affects Mac even though the code change doesn't seem like it should. 1410391 20 ross.kirsling 2018-03-29 14:10:09 -0700 At r230075 (after the rollout, running Xcode 9.2 on Sierra), I can confirm that: - `run-minibrowser` succeeds with or without the patch (both release and debug) - the "play" button in Xcode fails with or without the patch (WebProcess crash on launch) I'm not sure what's causing the issue in Xcode, but it's not this patch; I will plan to reland accordingly. 1410392 21 achristensen 2018-03-29 14:10:55 -0700 Woah, let Ryosuke double check before relanding. 1410393 22 achristensen 2018-03-29 14:12:29 -0700 (In reply to Ross Kirsling from comment #20) > At r230075 (after the rollout, running Xcode 9.2 on Sierra), I can confirm > that: What about r230069 before the rollout? 1410397 23 ross.kirsling 2018-03-29 14:15:50 -0700 (In reply to Alex Christensen from comment #21) > Woah, let Ryosuke double check before relanding. Whoops, okay. Standing by. (In reply to Alex Christensen from comment #22) > What about r230069 before the rollout? Not sure. Builds are a half-hour apiece so I've been working on this since 9:15 AM. 1410399 24 ross.kirsling 2018-03-29 14:24:11 -0700 (In reply to Ross Kirsling from comment #20) > - `run-minibrowser` succeeds with or without the patch (both release and > debug) Er, to be more clear, there's no reproduction with `build-webkit [--debug]` followed by `run-minibrowser [--debug]` or `run-safari [--debug]` on the command line. I don't really understand what differs between this and the build-and-run from within Xcode. 1410400 25 achristensen 2018-03-29 14:26:05 -0700 The definition of ALWAYS_INLINE depends on NDEBUG, so you'll need to do a release build. 1410401 26 ross.kirsling 2018-03-29 14:27:50 -0700 (In reply to Alex Christensen from comment #25) > The definition of ALWAYS_INLINE depends on NDEBUG, so you'll need to do a > release build. Sure, I did both across the board to be extra careful. 1410421 27 ross.kirsling 2018-03-29 15:28:51 -0700 Alright, one more scenario just to make things perfectly unambiguous. :D 1. git checkout dd7a49d554b8583f61cfda3f0a42ecdf3ed5b6de # i.e. r230061, the revision immediately before this patch landed 2. git clean -ffdx # because why not 3. Open Xcode, Product > Clean 4. Hit the Play button => WebContent process crashes immediately. 1410431 28 ross.kirsling 2018-03-29 16:12:35 -0700 (In reply to Ross Kirsling from comment #27) > Alright, one more scenario just to make things perfectly unambiguous. :D > > 1. git checkout dd7a49d554b8583f61cfda3f0a42ecdf3ed5b6de # i.e. r230061, the > revision immediately before this patch landed > > 2. git clean -ffdx # because why not > > 3. Open Xcode, Product > Clean > > 4. Hit the Play button > > => WebContent process crashes immediately. ...or apparently this doesn't count as a thorough clean, from Xcode's perspective. Evidently there's a file: /Users/.../Library/Developer/Xcode/DerivedData/WebKit-.../Build/Products/Debug/WebKit.framework/Resources/com.apple.WebProcess.sb To which my comment from this patch was added: > // Sometimes needed due to MSVC optimizer sensitivities. And that's syntactically invalid amid s-expressions. This leaves me with many questions, but I'm guessing that there's a reason why only block comments are used in WTF/Compiler.h. :( 1410434 29 achristensen 2018-03-29 16:14:23 -0700 Aha! Change the comment style and resubmit! 1410435 30 achristensen 2018-03-29 16:14:39 -0700 Actually, just remove the comment. 1410457 31 336823 ross.kirsling 2018-03-29 17:27:33 -0700 Created attachment 336823 Patch for landing 1410459 32 achristensen 2018-03-29 17:33:55 -0700 (In reply to Alex Christensen from comment #18) > (In reply to Ryosuke Niwa from comment #16) > This change makes no change on Mac. You have almost certainly misdiagnosed For the record, you did not misdiagnose, Ryosuke. 1410464 33 ross.kirsling 2018-03-29 17:48:48 -0700 (In reply to Alex Christensen from comment #32) > (In reply to Alex Christensen from comment #18) > > (In reply to Ryosuke Niwa from comment #16) > > This change makes no change on Mac. You have almost certainly misdiagnosed > For the record, you did not misdiagnose, Ryosuke. Indeed, what a trip. I've created https://bugs.webkit.org/show_bug.cgi?id=184166 to track the real bug here. :) 1410497 34 336823 commit-queue 2018-03-29 19:47:38 -0700 Comment on attachment 336823 Patch for landing Clearing flags on attachment: 336823 Committed r230091: <https://trac.webkit.org/changeset/230091> 1410498 35 commit-queue 2018-03-29 19:47:40 -0700 All reviewed patches have been landed. Closing bug. 336631 2018-03-27 17:17:34 -0700 2018-03-28 16:20:40 -0700 Patch bug-184062-20180327171744.patch text/plain 1229 ross.kirsling U3VidmVyc2lvbiBSZXZpc2lvbjogMjI5OTg5CmRpZmYgLS1naXQgYS9Tb3VyY2UvV1RGL0NoYW5n ZUxvZyBiL1NvdXJjZS9XVEYvQ2hhbmdlTG9nCmluZGV4IDlmNDQwZTM3MGY3YWMzNGM0MDNlZGZm YjA1NjE0Zjk0NWE4OGE0Y2MuLjQ4OGQ1YjM2ODEwMTcwMWEwMjkwNTYwM2QxZGNkNDBjODYxMWE0 ZDYgMTAwNjQ0Ci0tLSBhL1NvdXJjZS9XVEYvQ2hhbmdlTG9nCisrKyBiL1NvdXJjZS9XVEYvQ2hh bmdlTG9nCkBAIC0xLDMgKzEsMTMgQEAKKzIwMTgtMDMtMjcgIFJvc3MgS2lyc2xpbmcgIDxyb3Nz LmtpcnNsaW5nQHNvbnkuY29tPgorCisgICAgICAgIE1TVkMgX19mb3JjZWlubGluZSBzbG93cyBk b3duIEpTQyByZWxlYXNlIGJ1aWxkIGZpdmVmb2xkIGFmdGVyIHIyMjkzOTEKKyAgICAgICAgaHR0 cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTE4NDA2MgorCisgICAgICAgIFJl dmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgICogd3RmL0NvbXBpbGVyLmg6Cisg ICAgICAgIFVzZSBpbmxpbmUgYW5kIG5vdCBfX2ZvcmNlaW5saW5lIG9uIFdpbmRvd3MgdW50aWwg dGhlIE1TVkMgb3B0aW1pemVyIGNhbiBoYW5kbGUgaXQuCisKIDIwMTgtMDMtMjYgIFl1c3VrZSBT dXp1a2kgIDx1dGF0YW5lLnRlYUBnbWFpbC5jb20+CiAKICAgICAgICAgV2Ugc2hvdWxkIGhhdmUg U1NFNCBkZXRlY3Rpb24gaW4gdGhlIFg4NiBNYWNyb0Fzc2VtYmxlci4KZGlmZiAtLWdpdCBhL1Nv dXJjZS9XVEYvd3RmL0NvbXBpbGVyLmggYi9Tb3VyY2UvV1RGL3d0Zi9Db21waWxlci5oCmluZGV4 IGMxMGQ3OWM1N2NlNmJjYjI4NjE2ZDczODYyNGM0ZjM5MWZlODI3ZDMuLmY5NWQ1NzNjOWE2YzE5 MDNjNmE3NTkyNGFjNTFkMzYxMmZjODRkNWEgMTAwNjQ0Ci0tLSBhL1NvdXJjZS9XVEYvd3RmL0Nv bXBpbGVyLmgKKysrIGIvU291cmNlL1dURi93dGYvQ29tcGlsZXIuaApAQCAtMTcwLDEwICsxNzAs NiBAQAogI2RlZmluZSBBTFdBWVNfSU5MSU5FIGlubGluZSBfX2F0dHJpYnV0ZV9fKChfX2Fsd2F5 c19pbmxpbmVfXykpCiAjZW5kaWYKIAotI2lmICFkZWZpbmVkKEFMV0FZU19JTkxJTkUpICYmIENP TVBJTEVSKE1TVkMpICYmIGRlZmluZWQoTkRFQlVHKQotI2RlZmluZSBBTFdBWVNfSU5MSU5FIF9f Zm9yY2VpbmxpbmUKLSNlbmRpZgotCiAjaWYgIWRlZmluZWQoQUxXQVlTX0lOTElORSkKICNkZWZp bmUgQUxXQVlTX0lOTElORSBpbmxpbmUKICNlbmRpZgo= 336731 2018-03-28 16:20:42 -0700 2018-03-29 17:27:31 -0700 Patch bug-184062-20180328162053.patch text/plain 4959 ross.kirsling U3VidmVyc2lvbiBSZXZpc2lvbjogMjI5OTg5CmRpZmYgLS1naXQgYS9Tb3VyY2UvSmF2YVNjcmlw dENvcmUvQ2hhbmdlTG9nIGIvU291cmNlL0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwppbmRleCBh NjlmYTQ1Yjk0MmMwOWE2Y2I1ZjI3YjgyY2E0ZjIzY2VhM2YzYWUzLi4wZDc3NjdlY2YwMjk5OTM2 NjVkOGUxOTQ5YzdlNzU5MWE3MjY3NWEyIDEwMDY0NAotLS0gYS9Tb3VyY2UvSmF2YVNjcmlwdENv cmUvQ2hhbmdlTG9nCisrKyBiL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9DaGFuZ2VMb2cKQEAgLTEs MyArMSwxNCBAQAorMjAxOC0wMy0yOCAgUm9zcyBLaXJzbGluZyAgPHJvc3Mua2lyc2xpbmdAc29u eS5jb20+CisKKyAgICAgICAgTVNWQyBfX2ZvcmNlaW5saW5lIHNsb3dzIGRvd24gSlNDIHJlbGVh c2UgYnVpbGQgZml2ZWZvbGQgYWZ0ZXIgcjIyOTM5MQorICAgICAgICBodHRwczovL2J1Z3Mud2Vi a2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9MTg0MDYyCisKKyAgICAgICAgUmV2aWV3ZWQgYnkgTk9C T0RZIChPT1BTISkuCisKKyAgICAgICAgKiBqaXQvQ0NhbGxIZWxwZXJzLmg6CisgICAgICAgIChK U0M6OkNDYWxsSGVscGVyczo6bWFyc2hhbGxBcmd1bWVudFJlZ2lzdGVyKToKKyAgICAgICAgRXhl bXB0IE1TVkMgZnJvbSBhIHNpbmdsZSBmb3JjZWQgaW5saW5lIHVzZWQgd2l0aGluIHJlY3Vyc2l2 ZSB0ZW1wbGF0ZXMuCisKIDIwMTgtMDMtMjYgIFJvc3MgS2lyc2xpbmcgIDxyb3NzLmtpcnNsaW5n QHNvbnkuY29tPgogCiAgICAgICAgIEpJVCBjYWxsT3BlcmF0aW9uKCkgbmVlZHMgdG8gc3VwcG9y dCBvcGVyYXRpb25zIHRoYXQgcmV0dXJuIFNsb3dQYXRoUmV0dXJuVHlwZSBkaWZmZXJlbnRseSBv biBXaW5kb3dzLgpkaWZmIC0tZ2l0IGEvU291cmNlL1dURi9DaGFuZ2VMb2cgYi9Tb3VyY2UvV1RG L0NoYW5nZUxvZwppbmRleCA5ZjQ0MGUzNzBmN2FjMzRjNDAzZWRmZmIwNTYxNGY5NDVhODhhNGNj Li41MDYyZjY1MDE4NzQxZmE3YzljM2VjYzI5ZjA2YjBhMGU3ZWQ2NmFlIDEwMDY0NAotLS0gYS9T b3VyY2UvV1RGL0NoYW5nZUxvZworKysgYi9Tb3VyY2UvV1RGL0NoYW5nZUxvZwpAQCAtMSwzICsx LDEzIEBACisyMDE4LTAzLTI4ICBSb3NzIEtpcnNsaW5nICA8cm9zcy5raXJzbGluZ0Bzb255LmNv bT4KKworICAgICAgICBNU1ZDIF9fZm9yY2VpbmxpbmUgc2xvd3MgZG93biBKU0MgcmVsZWFzZSBi dWlsZCBmaXZlZm9sZCBhZnRlciByMjI5MzkxCisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQu b3JnL3Nob3dfYnVnLmNnaT9pZD0xODQwNjIKKworICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkg KE9PUFMhKS4KKworICAgICAgICAqIHd0Zi9Db21waWxlci5oOgorICAgICAgICBBZGQgQUxXQVlT X0lOTElORV9FWENFUFRfTVNWQyB0byBzdXBwb3J0IE1TVkMgb3B0aW1pemVyIHNlbnNpdGl2aXRp ZXMuCisKIDIwMTgtMDMtMjYgIFl1c3VrZSBTdXp1a2kgIDx1dGF0YW5lLnRlYUBnbWFpbC5jb20+ CiAKICAgICAgICAgV2Ugc2hvdWxkIGhhdmUgU1NFNCBkZXRlY3Rpb24gaW4gdGhlIFg4NiBNYWNy b0Fzc2VtYmxlci4KZGlmZiAtLWdpdCBhL1NvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZyBiL1NvdXJj ZS9XZWJDb3JlL0NoYW5nZUxvZwppbmRleCAxZTNiYjhkZDFmMDkzZDY3NDEzMmE1Y2NkMTU0NjNh MGMwOGIyYmIwLi5iZTQ2NWE1NDA3ODE3NWNiMWQ2MzY4NzI5NzJkODJkYmZiMGFiOTU4IDEwMDY0 NAotLS0gYS9Tb3VyY2UvV2ViQ29yZS9DaGFuZ2VMb2cKKysrIGIvU291cmNlL1dlYkNvcmUvQ2hh bmdlTG9nCkBAIC0xLDMgKzEsMTMgQEAKKzIwMTgtMDMtMjggIFJvc3MgS2lyc2xpbmcgIDxyb3Nz LmtpcnNsaW5nQHNvbnkuY29tPgorCisgICAgICAgIE1TVkMgX19mb3JjZWlubGluZSBzbG93cyBk b3duIEpTQyByZWxlYXNlIGJ1aWxkIGZpdmVmb2xkIGFmdGVyIHIyMjkzOTEKKyAgICAgICAgaHR0 cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTE4NDA2MgorCisgICAgICAgIFJl dmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgICogcGxhdGZvcm0vZ3JhcGhpY3Mv Rm9ybWF0Q29udmVydGVyLmNwcDoKKyAgICAgICAgRmFjdG9yIEFMV0FZU19JTkxJTkVfRVhDRVBU X01TVkMgb3V0IHRvIFdURi4KKwogMjAxOC0wMy0yMyAgQW50b2luZSBRdWludCAgPGdyYW91dHNA YXBwbGUuY29tPgogCiAgICAgICAgIFtXZWIgQW5pbWF0aW9uc10gTWFrZSBpbXBvcnRlZC9tb3pp bGxhL2Nzcy1hbmltYXRpb25zL3Rlc3RfYW5pbWF0aW9uLWN1cnJlbnR0aW1lLmh0bWwgcGFzcyBy ZWxpYWJseQpkaWZmIC0tZ2l0IGEvU291cmNlL0phdmFTY3JpcHRDb3JlL2ppdC9DQ2FsbEhlbHBl cnMuaCBiL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9qaXQvQ0NhbGxIZWxwZXJzLmgKaW5kZXggZDhk N2IzYTFkNGZmNmFjZDkxZTFjYjA5NWFmMDgzNjY3NDA4NGU5Zi4uY2I4NmVlODEzMGMwMGE3Yjc3 ODA1OWQ2MTI3M2Y3MjcxNGIwYzJkYyAxMDA2NDQKLS0tIGEvU291cmNlL0phdmFTY3JpcHRDb3Jl L2ppdC9DQ2FsbEhlbHBlcnMuaAorKysgYi9Tb3VyY2UvSmF2YVNjcmlwdENvcmUvaml0L0NDYWxs SGVscGVycy5oCkBAIC0yOTUsOCArMjk1LDkgQEAgcHJpdmF0ZToKIAogI2lmIFVTRShKU1ZBTFVF NjQpCiAKKyAgICAvLyBBdm9pZCBNU1ZDIG9wdGltaXphdGlvbiB0aW1lIGV4cGxvc2lvbiBhc3Nv Y2lhdGVkIHdpdGggX19mb3JjZWlubGluZSBpbiByZWN1cnNpdmUgdGVtcGxhdGVzLgogICAgIHRl bXBsYXRlPHR5cGVuYW1lIE9wZXJhdGlvblR5cGUsIHVuc2lnbmVkIG51bUdQUkFyZ3MsIHVuc2ln bmVkIG51bUdQUlNvdXJjZXMsIHVuc2lnbmVkIG51bUZQUkFyZ3MsIHVuc2lnbmVkIG51bUZQUlNv dXJjZXMsIHVuc2lnbmVkIGV4dHJhUG9rZSwgdHlwZW5hbWUgUmVnVHlwZSwgdHlwZW5hbWUuLi4g QXJncz4KLSAgICBBTFdBWVNfSU5MSU5FIHZvaWQgbWFyc2hhbGxBcmd1bWVudFJlZ2lzdGVyKEFy Z0NvbGxlY3Rpb248bnVtR1BSQXJncywgbnVtR1BSU291cmNlcywgbnVtRlBSQXJncywgbnVtRlBS U291cmNlcywgZXh0cmFQb2tlPiBhcmdTb3VyY2VSZWdzLCBSZWdUeXBlIGFyZywgQXJncy4uLiBh cmdzKQorICAgIEFMV0FZU19JTkxJTkVfRVhDRVBUX01TVkMgdm9pZCBtYXJzaGFsbEFyZ3VtZW50 UmVnaXN0ZXIoQXJnQ29sbGVjdGlvbjxudW1HUFJBcmdzLCBudW1HUFJTb3VyY2VzLCBudW1GUFJB cmdzLCBudW1GUFJTb3VyY2VzLCBleHRyYVBva2U+IGFyZ1NvdXJjZVJlZ3MsIFJlZ1R5cGUgYXJn LCBBcmdzLi4uIGFyZ3MpCiAgICAgewogICAgICAgICB1c2luZyBJbmZvVHlwZSA9IEluZm9UeXBl Rm9yUmVnPFJlZ1R5cGU+OwogICAgICAgICB1bnNpZ25lZCBudW1BcmdSZWdpc3RlcnMgPSBJbmZv VHlwZTo6bnVtYmVyT2ZBcmd1bWVudFJlZ2lzdGVyczsKZGlmZiAtLWdpdCBhL1NvdXJjZS9XVEYv d3RmL0NvbXBpbGVyLmggYi9Tb3VyY2UvV1RGL3d0Zi9Db21waWxlci5oCmluZGV4IGMxMGQ3OWM1 N2NlNmJjYjI4NjE2ZDczODYyNGM0ZjM5MWZlODI3ZDMuLjljYTdmNmU0MzkzNzZlMmNiNGZmNGU5 YmQ3YzJiYWE3MWE5YmRiNzIgMTAwNjQ0Ci0tLSBhL1NvdXJjZS9XVEYvd3RmL0NvbXBpbGVyLmgK KysrIGIvU291cmNlL1dURi93dGYvQ29tcGlsZXIuaApAQCAtMTc4LDYgKzE3OCwxMyBAQAogI2Rl ZmluZSBBTFdBWVNfSU5MSU5FIGlubGluZQogI2VuZGlmCiAKKy8vIFNvbWV0aW1lcyBuZWVkZWQg ZHVlIHRvIE1TVkMgb3B0aW1pemVyIHNlbnNpdGl2aXRpZXMuCisjaWYgQ09NUElMRVIoTVNWQykK KyNkZWZpbmUgQUxXQVlTX0lOTElORV9FWENFUFRfTVNWQyBpbmxpbmUKKyNlbHNlCisjZGVmaW5l IEFMV0FZU19JTkxJTkVfRVhDRVBUX01TVkMgQUxXQVlTX0lOTElORQorI2VuZGlmCisKIC8qIFdU Rl9FWFRFUk5fQ197QkVHSU4sIEVORH0gKi8KIAogI2lmZGVmIF9fY3BsdXNwbHVzCmRpZmYgLS1n aXQgYS9Tb3VyY2UvV2ViQ29yZS9wbGF0Zm9ybS9ncmFwaGljcy9Gb3JtYXRDb252ZXJ0ZXIuY3Bw IGIvU291cmNlL1dlYkNvcmUvcGxhdGZvcm0vZ3JhcGhpY3MvRm9ybWF0Q29udmVydGVyLmNwcApp bmRleCA3ZjAyNjM3NDA4ZjI5OTJjYzBlNmY2NWE1NmJiNGRkNDdmMTIzMzlkLi4zZmI5YTdlNGE2 YjAyZDE0NTVkYTc2Y2RiZTQwNjFkN2U4MDNhN2U3IDEwMDY0NAotLS0gYS9Tb3VyY2UvV2ViQ29y ZS9wbGF0Zm9ybS9ncmFwaGljcy9Gb3JtYXRDb252ZXJ0ZXIuY3BwCisrKyBiL1NvdXJjZS9XZWJD b3JlL3BsYXRmb3JtL2dyYXBoaWNzL0Zvcm1hdENvbnZlcnRlci5jcHAKQEAgLTM1LDEzICszNSw2 IEBACiAjaW5jbHVkZSAiR3JhcGhpY3NDb250ZXh0M0RORU9OLmgiCiAjZW5kaWYKIAotLy8gVmlz dWFsIFN0dWRpbyBjcmFzaGVzIHdpdGggYSBDMTA2MyBGYXRhbCBFcnJvciBpZiBldmVyeXRoaW5n IGlzIGlubGluZWQuCi0jaWYgQ09NUElMRVIoTVNWQykKLSNkZWZpbmUgQUxXQVlTX0lOTElORV9F WENFUFRfTVNWQwotI2Vsc2UKLSNkZWZpbmUgQUxXQVlTX0lOTElORV9FWENFUFRfTVNWQyBBTFdB WVNfSU5MSU5FCi0jZW5kaWYKLQogbmFtZXNwYWNlIFdlYkNvcmUgewogCiAKQEAgLTEyMDUsNiAr MTE5OCw3IEBAIEFMV0FZU19JTkxJTkUgdm9pZCBGb3JtYXRDb252ZXJ0ZXI6OmNvbnZlcnQoR3Jh cGhpY3NDb250ZXh0M0Q6OkFscGhhT3AgYWxwaGFPcCkKICN1bmRlZiBGT1JNQVRDT05WRVJURVJf Q0FTRV9BTFBIQU9QCiB9CiAKKy8vIFZpc3VhbCBTdHVkaW8gY3Jhc2hlcyB3aXRoIGEgQzEwNjMg RmF0YWwgRXJyb3IgaWYgZXZlcnl0aGluZyBpcyBpbmxpbmVkLgogdGVtcGxhdGU8R3JhcGhpY3ND b250ZXh0M0Q6OkRhdGFGb3JtYXQgU3JjRm9ybWF0LCBHcmFwaGljc0NvbnRleHQzRDo6RGF0YUZv cm1hdCBEc3RGb3JtYXQsIEdyYXBoaWNzQ29udGV4dDNEOjpBbHBoYU9wIGFscGhhT3A+CiBBTFdB WVNfSU5MSU5FX0VYQ0VQVF9NU1ZDIHZvaWQgRm9ybWF0Q29udmVydGVyOjpjb252ZXJ0KCkKIHsK 336823 2018-03-29 17:27:33 -0700 2018-03-29 19:47:38 -0700 Patch for landing bug-184062-20180329172732.patch text/plain 4856 ross.kirsling U3VidmVyc2lvbiBSZXZpc2lvbjogMjMwMDg4CmRpZmYgLS1naXQgYS9Tb3VyY2UvSmF2YVNjcmlw dENvcmUvQ2hhbmdlTG9nIGIvU291cmNlL0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwppbmRleCA0 ZGIzNGI0MzcwZjE4N2U4YjZhOTk2NjJkZjM1YzYxZTZhMmMzOTYwLi5mNjUyMzg0YzVjZmQxMWMy M2NiYTJjM2UyOTA0ODA3ZWVmZTc0NDFmIDEwMDY0NAotLS0gYS9Tb3VyY2UvSmF2YVNjcmlwdENv cmUvQ2hhbmdlTG9nCisrKyBiL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9DaGFuZ2VMb2cKQEAgLTEs MyArMSwxNCBAQAorMjAxOC0wMy0yOSAgUm9zcyBLaXJzbGluZyAgPHJvc3Mua2lyc2xpbmdAc29u eS5jb20+CisKKyAgICAgICAgTVNWQyBfX2ZvcmNlaW5saW5lIHNsb3dzIGRvd24gSlNDIHJlbGVh c2UgYnVpbGQgZml2ZWZvbGQgYWZ0ZXIgcjIyOTM5MQorICAgICAgICBodHRwczovL2J1Z3Mud2Vi a2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9MTg0MDYyCisKKyAgICAgICAgUmV2aWV3ZWQgYnkgQWxl eCBDaHJpc3RlbnNlbi4KKworICAgICAgICAqIGppdC9DQ2FsbEhlbHBlcnMuaDoKKyAgICAgICAg KEpTQzo6Q0NhbGxIZWxwZXJzOjptYXJzaGFsbEFyZ3VtZW50UmVnaXN0ZXIpOgorICAgICAgICBF eGVtcHQgTVNWQyBmcm9tIGEgc2luZ2xlIGZvcmNlZCBpbmxpbmUgdXNlZCB3aXRoaW4gcmVjdXJz aXZlIHRlbXBsYXRlcy4KKwogMjAxOC0wMy0yOSAgS2VpdGggTWlsbGVyICA8a2VpdGhfbWlsbGVy QGFwcGxlLmNvbT4KIAogICAgICAgICBBcnJheU1vZGUgc2hvdWxkIG5vdCB0cnkgdG8gZ2V0IHRo ZSBERkcgdG8gdGhpbmsgaXQgY2FuIGNvbnZlcnQgVHlwZWRBcnJheXMKZGlmZiAtLWdpdCBhL1Nv dXJjZS9XVEYvQ2hhbmdlTG9nIGIvU291cmNlL1dURi9DaGFuZ2VMb2cKaW5kZXggMjBlM2RmODJk MDgxYTVjODE0ZmEzYmM3ZjZjYzYwZDgyMjAxMGRlNi4uZTJjZDJiZDhmZjBkYzA4MDA4OTdjZjg1 YTQwMmFmZTM5ODBiY2FiZiAxMDA2NDQKLS0tIGEvU291cmNlL1dURi9DaGFuZ2VMb2cKKysrIGIv U291cmNlL1dURi9DaGFuZ2VMb2cKQEAgLTEsMyArMSwxMyBAQAorMjAxOC0wMy0yOSAgUm9zcyBL aXJzbGluZyAgPHJvc3Mua2lyc2xpbmdAc29ueS5jb20+CisKKyAgICAgICAgTVNWQyBfX2ZvcmNl aW5saW5lIHNsb3dzIGRvd24gSlNDIHJlbGVhc2UgYnVpbGQgZml2ZWZvbGQgYWZ0ZXIgcjIyOTM5 MQorICAgICAgICBodHRwczovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9MTg0MDYy CisKKyAgICAgICAgUmV2aWV3ZWQgYnkgQWxleCBDaHJpc3RlbnNlbi4KKworICAgICAgICAqIHd0 Zi9Db21waWxlci5oOgorICAgICAgICBBZGQgQUxXQVlTX0lOTElORV9FWENFUFRfTVNWQyB0byBz dXBwb3J0IE1TVkMgb3B0aW1pemVyIHNlbnNpdGl2aXRpZXMuCisKIDIwMTgtMDMtMjkgIENvbW1p dCBRdWV1ZSAgPGNvbW1pdC1xdWV1ZUB3ZWJraXQub3JnPgogCiAgICAgICAgIFVucmV2aWV3ZWQs IHJvbGxpbmcgb3V0IHIyMzAwNjIuCmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9DaGFuZ2VM b2cgYi9Tb3VyY2UvV2ViQ29yZS9DaGFuZ2VMb2cKaW5kZXggMWJlNjhmNGMwZmM0YjUyN2RmMzY3 YTc5MTQzZjA2YTJlN2Y2ZTkxMy4uYmE0YjM0MzhkMDA0MDQ1NzljOTRjODgzMmY5NDE4MGRkOGRi YTI0ZSAxMDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCisrKyBiL1NvdXJjZS9X ZWJDb3JlL0NoYW5nZUxvZwpAQCAtMSwzICsxLDEzIEBACisyMDE4LTAzLTI5ICBSb3NzIEtpcnNs aW5nICA8cm9zcy5raXJzbGluZ0Bzb255LmNvbT4KKworICAgICAgICBNU1ZDIF9fZm9yY2Vpbmxp bmUgc2xvd3MgZG93biBKU0MgcmVsZWFzZSBidWlsZCBmaXZlZm9sZCBhZnRlciByMjI5MzkxCisg ICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0xODQwNjIKKwor ICAgICAgICBSZXZpZXdlZCBieSBBbGV4IENocmlzdGVuc2VuLgorCisgICAgICAgICogcGxhdGZv cm0vZ3JhcGhpY3MvRm9ybWF0Q29udmVydGVyLmNwcDoKKyAgICAgICAgRmFjdG9yIEFMV0FZU19J TkxJTkVfRVhDRVBUX01TVkMgb3V0IHRvIFdURi4KKwogMjAxOC0wMy0yOSAgV2Vuc29uIEhzaWVo ICA8d2Vuc29uX2hzaWVoQGFwcGxlLmNvbT4KIAogICAgICAgICBGcmFtZVNlbGVjdGlvbjo6YXBw ZWFyYW5jZVVwZGF0ZVRpbWVyRmlyZWQgc2hvdWxkIGJlIHJvYnVzdCBhZ2FpbnN0IGxheW91dCBw YXNzZXMgdW5kZXJuZWF0aCBpdApkaWZmIC0tZ2l0IGEvU291cmNlL0phdmFTY3JpcHRDb3JlL2pp dC9DQ2FsbEhlbHBlcnMuaCBiL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9qaXQvQ0NhbGxIZWxwZXJz LmgKaW5kZXggZDhkN2IzYTFkNGZmNmFjZDkxZTFjYjA5NWFmMDgzNjY3NDA4NGU5Zi4uY2I4NmVl ODEzMGMwMGE3Yjc3ODA1OWQ2MTI3M2Y3MjcxNGIwYzJkYyAxMDA2NDQKLS0tIGEvU291cmNlL0ph dmFTY3JpcHRDb3JlL2ppdC9DQ2FsbEhlbHBlcnMuaAorKysgYi9Tb3VyY2UvSmF2YVNjcmlwdENv cmUvaml0L0NDYWxsSGVscGVycy5oCkBAIC0yOTUsOCArMjk1LDkgQEAgcHJpdmF0ZToKIAogI2lm IFVTRShKU1ZBTFVFNjQpCiAKKyAgICAvLyBBdm9pZCBNU1ZDIG9wdGltaXphdGlvbiB0aW1lIGV4 cGxvc2lvbiBhc3NvY2lhdGVkIHdpdGggX19mb3JjZWlubGluZSBpbiByZWN1cnNpdmUgdGVtcGxh dGVzLgogICAgIHRlbXBsYXRlPHR5cGVuYW1lIE9wZXJhdGlvblR5cGUsIHVuc2lnbmVkIG51bUdQ UkFyZ3MsIHVuc2lnbmVkIG51bUdQUlNvdXJjZXMsIHVuc2lnbmVkIG51bUZQUkFyZ3MsIHVuc2ln bmVkIG51bUZQUlNvdXJjZXMsIHVuc2lnbmVkIGV4dHJhUG9rZSwgdHlwZW5hbWUgUmVnVHlwZSwg dHlwZW5hbWUuLi4gQXJncz4KLSAgICBBTFdBWVNfSU5MSU5FIHZvaWQgbWFyc2hhbGxBcmd1bWVu dFJlZ2lzdGVyKEFyZ0NvbGxlY3Rpb248bnVtR1BSQXJncywgbnVtR1BSU291cmNlcywgbnVtRlBS QXJncywgbnVtRlBSU291cmNlcywgZXh0cmFQb2tlPiBhcmdTb3VyY2VSZWdzLCBSZWdUeXBlIGFy ZywgQXJncy4uLiBhcmdzKQorICAgIEFMV0FZU19JTkxJTkVfRVhDRVBUX01TVkMgdm9pZCBtYXJz aGFsbEFyZ3VtZW50UmVnaXN0ZXIoQXJnQ29sbGVjdGlvbjxudW1HUFJBcmdzLCBudW1HUFJTb3Vy Y2VzLCBudW1GUFJBcmdzLCBudW1GUFJTb3VyY2VzLCBleHRyYVBva2U+IGFyZ1NvdXJjZVJlZ3Ms IFJlZ1R5cGUgYXJnLCBBcmdzLi4uIGFyZ3MpCiAgICAgewogICAgICAgICB1c2luZyBJbmZvVHlw ZSA9IEluZm9UeXBlRm9yUmVnPFJlZ1R5cGU+OwogICAgICAgICB1bnNpZ25lZCBudW1BcmdSZWdp c3RlcnMgPSBJbmZvVHlwZTo6bnVtYmVyT2ZBcmd1bWVudFJlZ2lzdGVyczsKZGlmZiAtLWdpdCBh L1NvdXJjZS9XVEYvd3RmL0NvbXBpbGVyLmggYi9Tb3VyY2UvV1RGL3d0Zi9Db21waWxlci5oCmlu ZGV4IGMxMGQ3OWM1N2NlNmJjYjI4NjE2ZDczODYyNGM0ZjM5MWZlODI3ZDMuLjMzYjgwMzhhOTI2 YTkwMzM3ZWRiN2MyMTAyNmU1YjI4ZDNmNzJhZjcgMTAwNjQ0Ci0tLSBhL1NvdXJjZS9XVEYvd3Rm L0NvbXBpbGVyLmgKKysrIGIvU291cmNlL1dURi93dGYvQ29tcGlsZXIuaApAQCAtMTc4LDYgKzE3 OCwxMiBAQAogI2RlZmluZSBBTFdBWVNfSU5MSU5FIGlubGluZQogI2VuZGlmCiAKKyNpZiBDT01Q SUxFUihNU1ZDKQorI2RlZmluZSBBTFdBWVNfSU5MSU5FX0VYQ0VQVF9NU1ZDIGlubGluZQorI2Vs c2UKKyNkZWZpbmUgQUxXQVlTX0lOTElORV9FWENFUFRfTVNWQyBBTFdBWVNfSU5MSU5FCisjZW5k aWYKKwogLyogV1RGX0VYVEVSTl9DX3tCRUdJTiwgRU5EfSAqLwogCiAjaWZkZWYgX19jcGx1c3Bs dXMKZGlmZiAtLWdpdCBhL1NvdXJjZS9XZWJDb3JlL3BsYXRmb3JtL2dyYXBoaWNzL0Zvcm1hdENv bnZlcnRlci5jcHAgYi9Tb3VyY2UvV2ViQ29yZS9wbGF0Zm9ybS9ncmFwaGljcy9Gb3JtYXRDb252 ZXJ0ZXIuY3BwCmluZGV4IDdmMDI2Mzc0MDhmMjk5MmNjMGU2ZjY1YTU2YmI0ZGQ0N2YxMjMzOWQu LjNmYjlhN2U0YTZiMDJkMTQ1NWRhNzZjZGJlNDA2MWQ3ZTgwM2E3ZTcgMTAwNjQ0Ci0tLSBhL1Nv dXJjZS9XZWJDb3JlL3BsYXRmb3JtL2dyYXBoaWNzL0Zvcm1hdENvbnZlcnRlci5jcHAKKysrIGIv U291cmNlL1dlYkNvcmUvcGxhdGZvcm0vZ3JhcGhpY3MvRm9ybWF0Q29udmVydGVyLmNwcApAQCAt MzUsMTMgKzM1LDYgQEAKICNpbmNsdWRlICJHcmFwaGljc0NvbnRleHQzRE5FT04uaCIKICNlbmRp ZgogCi0vLyBWaXN1YWwgU3R1ZGlvIGNyYXNoZXMgd2l0aCBhIEMxMDYzIEZhdGFsIEVycm9yIGlm IGV2ZXJ5dGhpbmcgaXMgaW5saW5lZC4KLSNpZiBDT01QSUxFUihNU1ZDKQotI2RlZmluZSBBTFdB WVNfSU5MSU5FX0VYQ0VQVF9NU1ZDCi0jZWxzZQotI2RlZmluZSBBTFdBWVNfSU5MSU5FX0VYQ0VQ VF9NU1ZDIEFMV0FZU19JTkxJTkUKLSNlbmRpZgotCiBuYW1lc3BhY2UgV2ViQ29yZSB7CiAKIApA QCAtMTIwNSw2ICsxMTk4LDcgQEAgQUxXQVlTX0lOTElORSB2b2lkIEZvcm1hdENvbnZlcnRlcjo6 Y29udmVydChHcmFwaGljc0NvbnRleHQzRDo6QWxwaGFPcCBhbHBoYU9wKQogI3VuZGVmIEZPUk1B VENPTlZFUlRFUl9DQVNFX0FMUEhBT1AKIH0KIAorLy8gVmlzdWFsIFN0dWRpbyBjcmFzaGVzIHdp dGggYSBDMTA2MyBGYXRhbCBFcnJvciBpZiBldmVyeXRoaW5nIGlzIGlubGluZWQuCiB0ZW1wbGF0 ZTxHcmFwaGljc0NvbnRleHQzRDo6RGF0YUZvcm1hdCBTcmNGb3JtYXQsIEdyYXBoaWNzQ29udGV4 dDNEOjpEYXRhRm9ybWF0IERzdEZvcm1hdCwgR3JhcGhpY3NDb250ZXh0M0Q6OkFscGhhT3AgYWxw aGFPcD4KIEFMV0FZU19JTkxJTkVfRVhDRVBUX01TVkMgdm9pZCBGb3JtYXRDb252ZXJ0ZXI6OmNv bnZlcnQoKQogewo=