150794 2015-11-02 04:54:47 -0800 Fix the ENABLE(B3_JIT) build on Linux 2016-01-28 14:11:21 -0800 1 1 1 Unclassified WebKit JavaScriptCore Other Unspecified Unspecified RESOLVED FIXED P2 Normal --- 150279 152248 1 ossy ossy achristensen commit-queue fpizlo ggaren ossy ysuzuki oldest_to_newest 1138522 0 ossy 2015-11-02 04:54:47 -0800 SSIA 1138523 1 264574 ossy 2015-11-02 04:58:36 -0800 Created attachment 264574 Patch 1138524 2 264574 ossy 2015-11-02 05:17:14 -0800 Comment on attachment 264574 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=264574&action=review > Source/JavaScriptCore/b3/B3IndexSet.h:89 > iterator& operator++() > { > - m_iter++; > + ++m_iter; > return *this; > } The build failure which this change fixes: ../../Source/JavaScriptCore/b3/B3IndexSet.h: In member function ‘JSC::B3::IndexSet<T>::Iterable<CollectionType>::iterator& JSC::B3::IndexSet<T>::Iterable<CollectionType>::iterator::operator++()’: ../../Source/JavaScriptCore/b3/B3IndexSet.h:87:23: error: no ‘operator++(int)’ declared for postfix ‘++’ [-fpermissive] m_iter++; ^ I'm wondering how does it build on Apple Mac, because there isn't postfix operator++ in wtf/BitVector.h : BitVector::SetBitsIterable::iterator. Additionally using postfix operator++ isn't necessary here, since we don't use the previous value of m_iter. 1138526 3 264574 ossy 2015-11-02 05:22:11 -0800 Comment on attachment 264574 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=264574&action=review > Source/JavaScriptCore/b3/B3HeapRange.h:31 > +#include <limits.h> because we uses UINT_MAX in this file: ../../Source/JavaScriptCore/b3/B3HeapRange.h: In static member function 'static JSC::B3::HeapRange JSC::B3::HeapRange::top()': ../../Source/JavaScriptCore/b3/B3HeapRange.h:62:29: error: 'UINT_MAX' was not declared in this scope > Source/JavaScriptCore/b3/air/AirSpecial.cpp:31 > +#include <limits.h> ditto 1138527 4 264574 ossy 2015-11-02 05:26:57 -0800 Comment on attachment 264574 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=264574&action=review > Source/JavaScriptCore/b3/B3Type.h:73 > + RELEASE_ASSERT_NOT_REACHED(); > + return 0; > } The build failure which this change fixes: ../../Source/JavaScriptCore/b3/B3Type.h: In function 'size_t JSC::B3::sizeofType(JSC::B3::Type)': ../../Source/JavaScriptCore/b3/B3Type.h:71:1: error: control reaches end of non-void function [-Werror=return-type] cc1plus: all warnings being treated as errors Type isn't strongly typed enum, so we have to add this unreachable return to make the compiler happy. > Source/JavaScriptCore/b3/air/AirArg.h:127 > + RELEASE_ASSERT_NOT_REACHED(); > + return false; > } ditto > Source/JavaScriptCore/b3/air/AirArg.h:142 > + RELEASE_ASSERT_NOT_REACHED(); > + return false; > } ditto > Source/JavaScriptCore/b3/air/AirArg.h:477 > + RELEASE_ASSERT_NOT_REACHED(); > + return false; > } ditto > Source/JavaScriptCore/b3/air/AirArg.h:501 > + RELEASE_ASSERT_NOT_REACHED(); > + return false; > } ditto > Source/JavaScriptCore/b3/air/AirArg.h:531 > + RELEASE_ASSERT_NOT_REACHED(); > + return false; > } ditto > Source/JavaScriptCore/b3/air/AirCode.h:76 > + RELEASE_ASSERT_NOT_REACHED(); > + return { }; > } ditto > Source/JavaScriptCore/b3/air/AirCode.h:88 > + RELEASE_ASSERT_NOT_REACHED(); > + return 0; > } ditto 1138928 5 264574 commit-queue 2015-11-03 07:51:29 -0800 Comment on attachment 264574 Patch Clearing flags on attachment: 264574 Committed r191952: <http://trac.webkit.org/changeset/191952> 1138929 6 commit-queue 2015-11-03 07:51:33 -0800 All reviewed patches have been landed. Closing bug. 1138953 7 264574 fpizlo 2015-11-03 10:05:34 -0800 Comment on attachment 264574 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=264574&action=review >> Source/JavaScriptCore/b3/air/AirArg.h:127 >> } > > ditto Can we please find some other way of doing these kinds of fixes? These methods are called from hot paths, and on clang, this change causes the compiler to emit substantially more code. I don't like it when fixes for Linux unnecessarily slow things down on OSX. 1138955 8 264574 ossy 2015-11-03 10:15:34 -0800 Comment on attachment 264574 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=264574&action=review >>> Source/JavaScriptCore/b3/air/AirArg.h:127 >>> } >> >> ditto > > Can we please find some other way of doing these kinds of fixes? These methods are called from hot paths, and on clang, this change causes the compiler to emit substantially more code. I don't like it when fixes for Linux unnecessarily slow things down on OSX. Sure, I'll check if we can fix it in a different way. I agree that the code size can be a little bit bigger, but unreachable code can't slow down anything. 1138956 9 fpizlo 2015-11-03 10:21:04 -0800 (In reply to comment #8) > Comment on attachment 264574 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=264574&action=review > > >>> Source/JavaScriptCore/b3/air/AirArg.h:127 > >>> } > >> > >> ditto > > > > Can we please find some other way of doing these kinds of fixes? These methods are called from hot paths, and on clang, this change causes the compiler to emit substantially more code. I don't like it when fixes for Linux unnecessarily slow things down on OSX. > > Sure, I'll check if we can fix it in a different way. > > I agree that the code size can be a little bit bigger, but unreachable code > can't slow down anything. False. I just wrote this program, and compiled it with clang -O3. #include <stdlib.h> int foo(int x) { switch (x) { case 42: return 23; case 85: return 12; } } int bar(int x) { switch (x) { case 42: return 23; case 85: return 12; } return 65; } int baz(int x) { switch (x) { case 42: return 23; case 85: return 12; } abort(); return 65; } The output is: [pizlo@shakezilla OpenSource] otool -tv test.o test.o: (__TEXT,__text) section _foo: 0000000000000000 pushq %rbp 0000000000000001 movq %rsp, %rbp 0000000000000004 cmpl $0x55, %edi 0000000000000007 movl $0xc, %ecx 000000000000000c movl $0x17, %eax 0000000000000011 cmovel %ecx, %eax 0000000000000014 popq %rbp 0000000000000015 retq 0000000000000016 nopw %cs:(%rax,%rax) _bar: 0000000000000020 pushq %rbp 0000000000000021 movq %rsp, %rbp 0000000000000024 movl $0x17, %eax 0000000000000029 cmpl $0x2a, %edi 000000000000002c je 0x3f 000000000000002e cmpl $0x55, %edi 0000000000000031 jne 0x3a 0000000000000033 movl $0xc, %eax 0000000000000038 popq %rbp 0000000000000039 retq 000000000000003a movl $0x41, %eax 000000000000003f popq %rbp 0000000000000040 retq 0000000000000041 nopw %cs:(%rax,%rax) _baz: 0000000000000050 pushq %rbp 0000000000000051 movq %rsp, %rbp 0000000000000054 movl $0x17, %eax 0000000000000059 cmpl $0x2a, %edi 000000000000005c je 0x68 000000000000005e cmpl $0x55, %edi 0000000000000061 jne 0x6a 0000000000000063 movl $0xc, %eax 0000000000000068 popq %rbp 0000000000000069 retq 000000000000006a callq 0x6f As you can see, foo is compiled into a super-efficient conditional move. Bar and baz are both compiled into a pair of branches - you need to execute one branch to get to the first case, and two branches to get to the second case. Even one branch is worse than a conditional move, but two branches are a lot worse. It would actually be best if you reverted your change. There is no value to having B3 ported to Linux right now, and there is definitely harm to introducing any overheads, even hypothetical ones. We are working hard to turn B3 into a hotrod that generates code super quickly, and we'd rather not have any changes go in that have uncertain performance implications. 1138959 10 ossy 2015-11-03 10:22:55 -0800 Filed new bug report for follow up: bug150842 . I'll investigate it soon. 1138965 11 ossy 2015-11-03 10:34:01 -0800 (In reply to comment #9) > As you can see, foo is compiled into a super-efficient conditional move. > Bar and baz are both compiled into a pair of branches - you need to execute > one branch to get to the first case, and two branches to get to the second > case. Even one branch is worse than a conditional move, but two branches > are a lot worse. > It would actually be best if you reverted your change. There is no value to > having B3 ported to Linux right now, and there is definitely harm to > introducing any overheads, even hypothetical ones. We are working hard to > turn B3 into a hotrod that generates code super quickly, and we'd rather not > have any changes go in that have uncertain performance implications. OK, you convinced me, I reverted unreachable returns in http://trac.webkit.org/changeset/191957 . Will find a better fix bug150842 which doesn't generate any code. 1159862 12 ysuzuki 2016-01-28 13:45:35 -0800 Seeing https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62272, I noticed that we have very very easy (and no performance regression!!!) fix for this issue. Just replacing `addEdge` in the nested lambda to `this->addEdge`. Filip, how about this? 1159863 13 ysuzuki 2016-01-28 13:48:19 -0800 Ah, oops, I miscommented. https://bugs.webkit.org/show_bug.cgi?id=151624 is correct issue for me. 1159872 14 fpizlo 2016-01-28 14:11:21 -0800 (In reply to comment #12) > Seeing https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62272, I noticed that we > have very very easy (and no performance regression!!!) fix for this issue. > Just replacing `addEdge` in the nested lambda to `this->addEdge`. > > Filip, how about this? Seems sensible! 264574 2015-11-02 04:58:36 -0800 2015-11-03 07:51:29 -0800 Patch bug-150794-20151102135740.patch text/plain 6607 ossy U3VidmVyc2lvbiBSZXZpc2lvbjogMTkxODg2CmRpZmYgLS1naXQgYS9Tb3VyY2UvSmF2YVNjcmlw dENvcmUvQ2hhbmdlTG9nIGIvU291cmNlL0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwppbmRleCAw MjdiNzE5MmRkNTU0NjY1MDZlMWI0YTkzNTU2ZDhlMTc1MGYxOWU1Li4wNGRiN2IzMzRiNDlkOWVm NGIyYTYxOTNjZmQ3MDAyZTBmY2E5MzIxIDEwMDY0NAotLS0gYS9Tb3VyY2UvSmF2YVNjcmlwdENv cmUvQ2hhbmdlTG9nCisrKyBiL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9DaGFuZ2VMb2cKQEAgLTEs MyArMSwyOSBAQAorMjAxNS0xMS0wMiAgQ3NhYmEgT3N6dHJvZ29uw6FjICA8b3NzeUB3ZWJraXQu b3JnPgorCisgICAgICAgIEZpeCB0aGUgRU5BQkxFKEIzX0pJVCkgYnVpbGQgb24gTGludXgKKyAg ICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTE1MDc5NAorCisg ICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgICogQ01ha2VMaXN0 cy50eHQ6CisgICAgICAgICogYjMvQjNIZWFwUmFuZ2UuaDoKKyAgICAgICAgKiBiMy9CM0luZGV4 U2V0Lmg6CisgICAgICAgIChKU0M6OkIzOjpJbmRleFNldDo6SXRlcmFibGU6Oml0ZXJhdG9yOjpv cGVyYXRvcisrKToKKyAgICAgICAgKiBiMy9CM1R5cGUuaDoKKyAgICAgICAgKEpTQzo6QjM6OnNp emVvZlR5cGUpOgorICAgICAgICAqIGIzL2Fpci9BaXJBcmcuY3BwOgorICAgICAgICAoSlNDOjpC Mzo6QWlyOjpBcmc6OmR1bXApOgorICAgICAgICAqIGIzL2Fpci9BaXJBcmcuaDoKKyAgICAgICAg KEpTQzo6QjM6OkFpcjo6QXJnOjppc1VzZSk6CisgICAgICAgIChKU0M6OkIzOjpBaXI6OkFyZzo6 aXNEZWYpOgorICAgICAgICAoSlNDOjpCMzo6QWlyOjpBcmc6OmlzR1ApOgorICAgICAgICAoSlND OjpCMzo6QWlyOjpBcmc6OmlzRlApOgorICAgICAgICAoSlNDOjpCMzo6QWlyOjpBcmc6OmlzVHlw ZSk6CisgICAgICAgICogYjMvYWlyL0FpckNvZGUuaDoKKyAgICAgICAgKEpTQzo6QjM6OkFpcjo6 Q29kZTo6bmV3VG1wKToKKyAgICAgICAgKEpTQzo6QjM6OkFpcjo6Q29kZTo6bnVtVG1wcyk6Cisg ICAgICAgICogYjMvYWlyL0FpclNwZWNpYWwuY3BwOgorCiAyMDE1LTExLTAxICBZdXN1a2UgU3V6 dWtpICA8dXRhdGFuZS50ZWFAZ21haWwuY29tPgogCiAgICAgICAgIFtFUzZdIFN1cHBvcnQgR2Vu ZXJhdG9yIFN5bnRheApkaWZmIC0tZ2l0IGEvU291cmNlL0phdmFTY3JpcHRDb3JlL0NNYWtlTGlz dHMudHh0IGIvU291cmNlL0phdmFTY3JpcHRDb3JlL0NNYWtlTGlzdHMudHh0CmluZGV4IGRmMzcz M2QxOGRmNDNmNTg2ZGRkNGI0M2Y4NGRhMzIwNDIzZTgxNDEuLjY0MzBlZWU2MDBlNWJkZWQ3ZGM0 ODZkYjZjYWI1OWJjNGYwODdhN2UgMTAwNjQ0Ci0tLSBhL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9D TWFrZUxpc3RzLnR4dAorKysgYi9Tb3VyY2UvSmF2YVNjcmlwdENvcmUvQ01ha2VMaXN0cy50eHQK QEAgLTczLDYgKzczLDcgQEAgc2V0KEphdmFTY3JpcHRDb3JlX1NPVVJDRVMKICAgICBiMy9haXIv QWlyQmFzaWNCbG9jay5jcHAKICAgICBiMy9haXIvQWlyQ0NhbGxTcGVjaWFsLmNwcAogICAgIGIz L2Fpci9BaXJDb2RlLmNwcAorICAgIGIzL2Fpci9BaXJFbGltaW5hdGVEZWFkQ29kZS5jcHAKICAg ICBiMy9haXIvQWlyR2VuZXJhdGUuY3BwCiAgICAgYjMvYWlyL0FpckdlbmVyYXRlZC5jcHAKICAg ICBiMy9haXIvQWlySGFuZGxlQ2FsbGVlU2F2ZXMuY3BwCkBAIC04MCw2ICs4MSw3IEBAIHNldChK YXZhU2NyaXB0Q29yZV9TT1VSQ0VTCiAgICAgYjMvYWlyL0Fpckluc3QuY3BwCiAgICAgYjMvYWly L0FpclBoYXNlU2NvcGUuY3BwCiAgICAgYjMvYWlyL0FpclJlZ2lzdGVyUHJpb3JpdHkuY3BwCisg ICAgYjMvYWlyL0FpclJlcG9ydFVzZWRSZWdpc3RlcnMuY3BwCiAgICAgYjMvYWlyL0FpclNwZWNp YWwuY3BwCiAgICAgYjMvYWlyL0FpclNwaWxsRXZlcnl0aGluZy5jcHAKICAgICBiMy9haXIvQWly U3RhY2tTbG90LmNwcApAQCAtMTE2LDYgKzExOCw3IEBAIHNldChKYXZhU2NyaXB0Q29yZV9TT1VS Q0VTCiAgICAgYjMvQjNTdGFja1Nsb3RWYWx1ZS5jcHAKICAgICBiMy9CM1N3aXRjaENhc2UuY3Bw CiAgICAgYjMvQjNTd2l0Y2hWYWx1ZS5jcHAKKyAgICBiMy9CM1RpbWluZ1Njb3BlLmNwcAogICAg IGIzL0IzVHlwZS5jcHAKICAgICBiMy9CM1Vwc2lsb25WYWx1ZS5jcHAKICAgICBiMy9CM1VzZUNv dW50cy5jcHAKZGlmZiAtLWdpdCBhL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9iMy9CM0hlYXBSYW5n ZS5oIGIvU291cmNlL0phdmFTY3JpcHRDb3JlL2IzL0IzSGVhcFJhbmdlLmgKaW5kZXggMTAwMWVj YzU5YzMxYTlhMTMxNTM2MDA2NzQ0NTk2MTUzOWQxMmNiMi4uNTFlZjE1ZGFmNjQ5ZWRkY2QyZDJl NDc1NGE0NzQwYzI1OTYzOTFiYSAxMDA2NDQKLS0tIGEvU291cmNlL0phdmFTY3JpcHRDb3JlL2Iz L0IzSGVhcFJhbmdlLmgKKysrIGIvU291cmNlL0phdmFTY3JpcHRDb3JlL2IzL0IzSGVhcFJhbmdl LmgKQEAgLTI4LDYgKzI4LDcgQEAKIAogI2lmIEVOQUJMRShCM19KSVQpCiAKKyNpbmNsdWRlIDxs aW1pdHMuaD4KICNpbmNsdWRlIDx3dGYvTWF0aEV4dHJhcy5oPgogI2luY2x1ZGUgPHd0Zi9Qcmlu dFN0cmVhbS5oPgogCmRpZmYgLS1naXQgYS9Tb3VyY2UvSmF2YVNjcmlwdENvcmUvYjMvQjNJbmRl eFNldC5oIGIvU291cmNlL0phdmFTY3JpcHRDb3JlL2IzL0IzSW5kZXhTZXQuaAppbmRleCA0MWQ3 MWQ4ZWU2MzExMjA2YmNkNzgwNGY4ODJiMzdhZTFkMGZkZGIzLi43NzJjNTExNzY5NTljNzA3NjUw NDk2YjQzZDI1NjMxZDY1NWEyZmI0IDEwMDY0NAotLS0gYS9Tb3VyY2UvSmF2YVNjcmlwdENvcmUv YjMvQjNJbmRleFNldC5oCisrKyBiL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9iMy9CM0luZGV4U2V0 LmgKQEAgLTg0LDcgKzg0LDcgQEAgcHVibGljOgogCiAgICAgICAgICAgICBpdGVyYXRvciYgb3Bl cmF0b3IrKygpCiAgICAgICAgICAgICB7Ci0gICAgICAgICAgICAgICAgbV9pdGVyKys7CisgICAg ICAgICAgICAgICAgKyttX2l0ZXI7CiAgICAgICAgICAgICAgICAgcmV0dXJuICp0aGlzOwogICAg ICAgICAgICAgfQogCmRpZmYgLS1naXQgYS9Tb3VyY2UvSmF2YVNjcmlwdENvcmUvYjMvQjNUeXBl LmggYi9Tb3VyY2UvSmF2YVNjcmlwdENvcmUvYjMvQjNUeXBlLmgKaW5kZXggY2Y0YTU2MDU1NzU1 OGJmNzU3ZTRjYzJjZDZkNWJmNzg0YzdhODE2My4uYTkxNDYwMTUzYzJmNTdiNThlMDg3YTUzZGM1 YmUyYmU3MGRmMTgyNyAxMDA2NDQKLS0tIGEvU291cmNlL0phdmFTY3JpcHRDb3JlL2IzL0IzVHlw ZS5oCisrKyBiL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9iMy9CM1R5cGUuaApAQCAtNjgsNiArNjgs OCBAQCBpbmxpbmUgc2l6ZV90IHNpemVvZlR5cGUoVHlwZSB0eXBlKQogICAgIGNhc2UgRG91Ymxl OgogICAgICAgICByZXR1cm4gODsKICAgICB9CisgICAgUkVMRUFTRV9BU1NFUlRfTk9UX1JFQUNI RUQoKTsKKyAgICByZXR1cm4gMDsKIH0KIAogfSB9IC8vIG5hbWVzcGFjZSBKU0M6OkIzCmRpZmYg LS1naXQgYS9Tb3VyY2UvSmF2YVNjcmlwdENvcmUvYjMvYWlyL0FpckFyZy5jcHAgYi9Tb3VyY2Uv SmF2YVNjcmlwdENvcmUvYjMvYWlyL0FpckFyZy5jcHAKaW5kZXggNThhMzA3ZGIzNmIzNGJjZDc5 MDIyZTdiZmY1NTc5ZTViM2VhZDU5OC4uN2FlODdmMzMxNmNlYzE1ZDkzZmQ0NWY4Y2Y0Yzc1ODVj NDRiYWZjZSAxMDA2NDQKLS0tIGEvU291cmNlL0phdmFTY3JpcHRDb3JlL2IzL2Fpci9BaXJBcmcu Y3BwCisrKyBiL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9iMy9haXIvQWlyQXJnLmNwcApAQCAtNDYs NyArNDYsNyBAQCB2b2lkIEFyZzo6ZHVtcChQcmludFN0cmVhbSYgb3V0KSBjb25zdAogICAgICAg ICBvdXQucHJpbnQoIiQiLCBtX29mZnNldCk7CiAgICAgICAgIHJldHVybjsKICAgICBjYXNlIElt bTY0OgotICAgICAgICBvdXQucHJpbnRmKCIkMHglbGx4IiwgbV9vZmZzZXQpOworICAgICAgICBv dXQucHJpbnRmKCIkMHglbGx4Iiwgc3RhdGljX2Nhc3Q8bG9uZyBsb25nIHVuc2lnbmVkPihtX29m ZnNldCkpOwogICAgICAgICByZXR1cm47CiAgICAgY2FzZSBBZGRyOgogICAgICAgICBpZiAob2Zm c2V0KCkpCmRpZmYgLS1naXQgYS9Tb3VyY2UvSmF2YVNjcmlwdENvcmUvYjMvYWlyL0FpckFyZy5o IGIvU291cmNlL0phdmFTY3JpcHRDb3JlL2IzL2Fpci9BaXJBcmcuaAppbmRleCA2YjlmY2E1MTA5 Njk0ZWViN2Y5MjI0ODM4MDJhNTJjOTYzNGExZGFiLi45OWM0NzNiNjRjNGFkNmVhM2ZhMGVhNGE4 ZGNlMjQ1MmQ3MDZlMzQxIDEwMDY0NAotLS0gYS9Tb3VyY2UvSmF2YVNjcmlwdENvcmUvYjMvYWly L0FpckFyZy5oCisrKyBiL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9iMy9haXIvQWlyQXJnLmgKQEAg LTEyMiw2ICsxMjIsOCBAQCBwdWJsaWM6CiAgICAgICAgIGNhc2UgVXNlQWRkcjoKICAgICAgICAg ICAgIHJldHVybiBmYWxzZTsKICAgICAgICAgfQorICAgICAgICBSRUxFQVNFX0FTU0VSVF9OT1Rf UkVBQ0hFRCgpOworICAgICAgICByZXR1cm4gZmFsc2U7CiAgICAgfQogCiAgICAgLy8gUmV0dXJu cyB0cnVlIGlmIHRoZSBSb2xlIGltcGxpZXMgdGhhdCB0aGUgSW5zdCB3aWxsIERlZiB0aGUgQXJn LgpAQCAtMTM1LDYgKzEzNyw4IEBAIHB1YmxpYzoKICAgICAgICAgY2FzZSBVc2VEZWY6CiAgICAg ICAgICAgICByZXR1cm4gdHJ1ZTsKICAgICAgICAgfQorICAgICAgICBSRUxFQVNFX0FTU0VSVF9O T1RfUkVBQ0hFRCgpOworICAgICAgICByZXR1cm4gZmFsc2U7CiAgICAgfQogCiAgICAgc3RhdGlj IFR5cGUgdHlwZUZvckIzVHlwZShCMzo6VHlwZSB0eXBlKQpAQCAtNDY4LDYgKzQ3Miw4IEBAIHB1 YmxpYzoKICAgICAgICAgY2FzZSBJbnZhbGlkOgogICAgICAgICAgICAgcmV0dXJuIGZhbHNlOwog ICAgICAgICB9CisgICAgICAgIFJFTEVBU0VfQVNTRVJUX05PVF9SRUFDSEVEKCk7CisgICAgICAg IHJldHVybiBmYWxzZTsKICAgICB9CiAKICAgICAvLyBUZWxscyB1cyBpZiB0aGlzIEFyZyBjYW4g YmUgdXNlZCBpbiBhIHBvc2l0aW9uIHRoYXQgcmVxdWlyZXMgYSBGUCB2YWx1ZS4KQEAgLTQ5MCw2 ICs0OTYsOCBAQCBwdWJsaWM6CiAgICAgICAgIGNhc2UgVG1wOgogICAgICAgICAgICAgcmV0dXJu IGlzRlBUbXAoKTsKICAgICAgICAgfQorICAgICAgICBSRUxFQVNFX0FTU0VSVF9OT1RfUkVBQ0hF RCgpOworICAgICAgICByZXR1cm4gZmFsc2U7CiAgICAgfQogCiAgICAgYm9vbCBoYXNUeXBlKCkg Y29uc3QKQEAgLTUxOCw2ICs1MjYsOCBAQCBwdWJsaWM6CiAgICAgICAgIGNhc2UgRlA6CiAgICAg ICAgICAgICByZXR1cm4gaXNGUCgpOwogICAgICAgICB9CisgICAgICAgIFJFTEVBU0VfQVNTRVJU X05PVF9SRUFDSEVEKCk7CisgICAgICAgIHJldHVybiBmYWxzZTsKICAgICB9CiAKICAgICBib29s IGlzR1BSKCkgY29uc3QKZGlmZiAtLWdpdCBhL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9iMy9haXIv QWlyQ29kZS5oIGIvU291cmNlL0phdmFTY3JpcHRDb3JlL2IzL2Fpci9BaXJDb2RlLmgKaW5kZXgg ZTc3NzRlZWQzODc3ODE4MGNkMDlhMzEzYjA0MDQxNmQwNjE5ZmFmMS4uOTZlMjg2YmVjNTBjZmU1 MTIzZDAwNWMzN2ZiOWQ4ZWUzMjBkYzZlNSAxMDA2NDQKLS0tIGEvU291cmNlL0phdmFTY3JpcHRD b3JlL2IzL2Fpci9BaXJDb2RlLmgKKysrIGIvU291cmNlL0phdmFTY3JpcHRDb3JlL2IzL2Fpci9B aXJDb2RlLmgKQEAgLTcxLDYgKzcxLDggQEAgcHVibGljOgogICAgICAgICBjYXNlIEFyZzo6RlA6 CiAgICAgICAgICAgICByZXR1cm4gVG1wOjpmcFRtcEZvckluZGV4KG1fbnVtRlBUbXBzKyspOwog ICAgICAgICB9CisgICAgICAgIFJFTEVBU0VfQVNTRVJUX05PVF9SRUFDSEVEKCk7CisgICAgICAg IHJldHVybiB7IH07CiAgICAgfQogCiAgICAgdW5zaWduZWQgbnVtVG1wcyhBcmc6OlR5cGUgdHlw ZSkKQEAgLTgxLDYgKzgzLDggQEAgcHVibGljOgogICAgICAgICBjYXNlIEFyZzo6RlA6CiAgICAg ICAgICAgICByZXR1cm4gbV9udW1GUFRtcHM7CiAgICAgICAgIH0KKyAgICAgICAgUkVMRUFTRV9B U1NFUlRfTk9UX1JFQUNIRUQoKTsKKyAgICAgICAgcmV0dXJuIDA7CiAgICAgfQogCiAgICAgdW5z aWduZWQgY2FsbEFyZ0FyZWFTaXplKCkgY29uc3QgeyByZXR1cm4gbV9jYWxsQXJnQXJlYVNpemU7 IH0KZGlmZiAtLWdpdCBhL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9iMy9haXIvQWlyU3BlY2lhbC5j cHAgYi9Tb3VyY2UvSmF2YVNjcmlwdENvcmUvYjMvYWlyL0FpclNwZWNpYWwuY3BwCmluZGV4IDYy NzA3NmJlNGQwYjEzMjQ3ODFiNzJlMWNjYzQxYjBlNzFhYWRiZjYuLjM5OTVkMjYwNjBlMGExMWUz ZDhkYWExZmU4OTE2OWFkOWFlODQyNDQgMTAwNjQ0Ci0tLSBhL1NvdXJjZS9KYXZhU2NyaXB0Q29y ZS9iMy9haXIvQWlyU3BlY2lhbC5jcHAKKysrIGIvU291cmNlL0phdmFTY3JpcHRDb3JlL2IzL2Fp ci9BaXJTcGVjaWFsLmNwcApAQCAtMjgsNiArMjgsNyBAQAogCiAjaWYgRU5BQkxFKEIzX0pJVCkK IAorI2luY2x1ZGUgPGxpbWl0cy5oPgogI2luY2x1ZGUgPHd0Zi9TdHJpbmdQcmludFN0cmVhbS5o PgogCiBuYW1lc3BhY2UgSlNDIHsgbmFtZXNwYWNlIEIzIHsgbmFtZXNwYWNlIEFpciB7Cg==