170720 2017-04-11 02:12:43 -0700 [JSC] Date.parse should accept wider range of representation 2017-07-25 12:46:45 -0700 1 1 1 Unclassified WebKit JavaScriptCore WebKit Nightly Build Unspecified Unspecified RESOLVED FIXED P2 Normal --- 1 ysuzuki ysuzuki annulen benjamin buildbot cdumez cmarcelo darin dbates fpizlo ggaren jfbastien keith_miller mark.lam msaboff saam sam simon.fraser vitaliy.slobodin oldest_to_newest 1296322 0 ysuzuki 2017-04-11 02:12:43 -0700 Date.parse now conforms the spec. But it is a bit strict and users want to parse some more relaxed ones[1]. What we would like is adding a fallback path to try to parse it if the current Date.parse fails to parse. [1]: https://github.com/ariya/phantomjs/issues/11151 1296988 1 306985 ysuzuki 2017-04-13 05:23:11 -0700 Created attachment 306985 Patch 1297043 2 306985 darin 2017-04-13 11:27:16 -0700 Comment on attachment 306985 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=306985&action=review > Source/WTF/wtf/DateMath.cpp:897 > + std::optional<int> year = std::nullopt; No need for "= std::nullopt". That’s what optionals do when not specified. > Source/WTF/wtf/DateMath.cpp:1132 > + // 1. According to the V8, this default value is compatible to the old KJS. While current V8 uses 2001, > + // SpiderMonkey does not have such a fallback path. So this reason is not big deal. I am a bit confused by this comment. V8 is the name of a JavaScript engine, not a person, so I don’t know what "according to the V8" means; maybe someone on the V8 team did some research about the behavior of older versions of WebKit? I don’t know what "compatible to the old KJS" means or why it’s relevant to our decision of what behavior is best. “The old KJS” seems quite unimportant; very few websites were tailored to assume the behavior of the KJS engine since it wasn’t used by all that many people. So maybe it’s simply a peculiar way to talk about older versions of WebKit and JavaScriptCore. The term “current V8” is not good since the comment is likely to last a long time and this may no longer be true in the future, so we should probably say this in a different way. Sometimes I use the phrase "as of this writing". Here’s a comment I might write, but I don’t know if it’s accurate: // 1. Year 2000 was used in older versions of JavaScriptCore, thus some older content // tested only with WebKit might rely on this behavior (according to research done by the V8 team). // (As of April 2017, V8 is using the year 2001 and Spider Monkey is not doing this kind of fallback.) But when written this way it becomes clear that it’s a little strange to rely on research done by the V8 team to know what the behavior of old versions of WebKit is. Can’t we just check this rather than relying on what they say? Also, it would be nice to be precise rather than just "older versions of JavaScriptCore". 1297353 3 306985 ysuzuki 2017-04-14 00:56:14 -0700 Comment on attachment 306985 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=306985&action=review >> Source/WTF/wtf/DateMath.cpp:1132 >> + // SpiderMonkey does not have such a fallback path. So this reason is not big deal. > > I am a bit confused by this comment. V8 is the name of a JavaScript engine, not a person, so I don’t know what "according to the V8" means; maybe someone on the V8 team did some research about the behavior of older versions of WebKit? I don’t know what "compatible to the old KJS" means or why it’s relevant to our decision of what behavior is best. “The old KJS” seems quite unimportant; very few websites were tailored to assume the behavior of the KJS engine since it wasn’t used by all that many people. So maybe it’s simply a peculiar way to talk about older versions of WebKit and JavaScriptCore. The term “current V8” is not good since the comment is likely to last a long time and this may no longer be true in the future, so we should probably say this in a different way. Sometimes I use the phrase "as of this writing". Here’s a comment I might write, but I don’t know if it’s accurate: > > // 1. Year 2000 was used in older versions of JavaScriptCore, thus some older content > // tested only with WebKit might rely on this behavior (according to research done by the V8 team). > // (As of April 2017, V8 is using the year 2001 and Spider Monkey is not doing this kind of fallback.) > > But when written this way it becomes clear that it’s a little strange to rely on research done by the V8 team to know what the behavior of old versions of WebKit is. Can’t we just check this rather than relying on what they say? > > Also, it would be nice to be precise rather than just "older versions of JavaScriptCore". I've dived into old logs... This function is added at r34872 and not changed so much. While year is first initialized 0 and later it will be converted to 2000, we do not produce any values without parsing year before this patch. So, users won't see 2000 default value I think. 1297356 4 ysuzuki 2017-04-14 01:00:04 -0700 Committed r215359: <http://trac.webkit.org/changeset/215359> 1331829 5 simon.fraser 2017-07-24 18:51:07 -0700 This caused the graphs on https://www.recode.net/2017/7/18/15994992/isabel-ge-mahe-apple-china-managing-director to render incorrectly. 1331880 6 darin 2017-07-24 22:24:49 -0700 Do we know why it caused that, specifically what date strings were being passed? 1331924 7 ysuzuki 2017-07-25 05:04:34 -0700 Oh, that's very interesting b/c it means some libs would depend on super tricky behavior. I'll setup the environment to check what is going. 1331936 8 simon.fraser 2017-07-25 08:35:22 -0700 I didn't have time to do any investigation other than autospading. 1332025 9 ysuzuki 2017-07-25 11:35:59 -0700 I've investigated it. ## Situation This web site passes a string like "Jun. ‘15" to `Date.parse`. To make this site work correctly, we need to return NaN for this value. This patch relaxes Date parsing, then, for the above case, we return 2000/Jun/15 Date. This is expected behavior. As you can see the phantomjs issue in this bug, webdevs want to generate ????/Jun/15 for a string like "Jun 15". Why Chrome works fine is that it returns NaN for "Jun. ‘15". But it returns 2001/Jun/15 for "Jun. 15", "Jun 15" etc. ## In terms of the spec ECMA262's guarantee about Date.parse is not-strict. ECMA262 requires that some descriptive format can be parsed by `Date.parse` correctly. In the case of the other values, Date.parse may either succeed or fail. It's implementation dependent heuristic. So, with regard to the spec, our current behavior is correct. And ideally, users should not assume that the value must be rejected by Date.parse because the spec allows us to parse it with some heuristic. But the real world is not ideal....... :( https://tc39.github.io/ecma262/#sec-date.parse ## Possible solutions 1. Revert this patch By reverting this patch, we return NaN for "Jun. ‘15". But instead, we lose the ability to return 2001/Jun/15 for "Jun. 15", "Jun 15" etc. which is requested in phantomjs's issue (And sometimes we see this request in twitter). 2. Add a hack for "Jun. ‘15" case Add an additional heuristic for "Jun. ‘15" not to parse it as Date. Possible thing is that, if we see some leading text before the date (15), we do not recognize it as date. But it would break some existing heuristic. 3. Implement the heuristic completely same to the V8's one Strictly speaking, these heuristics are not specified in the spec. So, the request in the phantomjs issue is basically saying "We want to parse Date as if V8 parses it". The simplest solution for that is importing the heuristics in V8. 4. Request the site to fix it Ideally, anybody should not assume that Date.parse should fail with a specific format. The heuristics could recognize it. To me, for the short term solution, (1) should be appropriate. (2) is a bit too dangerous. For the long term solution, I think we have (1), (3), and (4). If we do not want to relax it, (1) and leave the current heuristic as is is the solution. If we would like to accept some form of Date which is requested in the original issue, we have (3) or (4). If we need to fix it quickly right now, we should take (1). 1332028 10 simon.fraser 2017-07-25 11:39:58 -0700 How do Gecko and Edge's Date parsing compare? 1332030 11 ysuzuki 2017-07-25 11:48:39 -0700 (In reply to Simon Fraser (smfr) from comment #10) > How do Gecko and Edge's Date parsing compare? Because I don't have Windows machine in my office, I cannot check it in Edge right now (anyone has Edge machine?). Firefox has different heuristic mechanism from V8. And it returns NaN for "Jun. ‘15", "Jun 15", and "Jun. 15". 1332031 12 simon.fraser 2017-07-25 11:52:19 -0700 You can get an Edge VM at https://developer.microsoft.com/en-us/microsoft-edge/tools/vms/ 1332048 13 ysuzuki 2017-07-25 12:46:45 -0700 (In reply to Simon Fraser (smfr) from comment #12) > You can get an Edge VM at > https://developer.microsoft.com/en-us/microsoft-edge/tools/vms/ Thank you! I've just downloaded and tested. Edge shows the same behavior to Firefox. And it returns NaN for "Jun. ‘15", "Jun 15", and "Jun. 15". So, not relaxing our heuristic is one idea. Basically, phantom.js's issue is raised because it is different from V8. 306985 2017-04-13 05:23:11 -0700 2017-04-13 11:27:16 -0700 Patch bug-170720-20170413212310.patch text/plain 8267 ysuzuki U3VidmVyc2lvbiBSZXZpc2lvbjogMjE1MzE0CmRpZmYgLS1naXQgYS9Tb3VyY2UvV1RGL0NoYW5n ZUxvZyBiL1NvdXJjZS9XVEYvQ2hhbmdlTG9nCmluZGV4IDliNDFjMGI4MWQ3NmM1MWE5MzQ3NTVj M2E3ZDQ0NThkNWY3YTJlNjAuLjliNzgxNjA5ODhhYTE5MDJkM2UyNzQ2ODg4NGU3ZjJjYWYwYWNj NzggMTAwNjQ0Ci0tLSBhL1NvdXJjZS9XVEYvQ2hhbmdlTG9nCisrKyBiL1NvdXJjZS9XVEYvQ2hh bmdlTG9nCkBAIC0xLDMgKzEsMjQgQEAKKzIwMTctMDQtMTMgIFl1c3VrZSBTdXp1a2kgIDx1dGF0 YW5lLnRlYUBnbWFpbC5jb20+CisKKyAgICAgICAgW0pTQ10gRGF0ZS5wYXJzZSBzaG91bGQgYWNj ZXB0IHdpZGVyIHJhbmdlIG9mIHJlcHJlc2VudGF0aW9uCisgICAgICAgIGh0dHBzOi8vYnVncy53 ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0xNzA3MjAKKworICAgICAgICBSZXZpZXdlZCBieSBO T0JPRFkgKE9PUFMhKS4KKworICAgICAgICBXZSB3b3VsZCBsaWtlIHRvIHJlbGF4IERhdGUucGFy c2UgdG8gYWNjZXB0IHRoZSByZXByZXNlbnRhdGlvbiAiTWF5IDgiLgorICAgICAgICBBdCB0aGF0 IHRpbWUsIHdlIGNob29zZSBgMjAwMGAgYXMgYSBkZWZhdWx0IHllYXIuIFRoaXMgaXMgYmVjYXVz ZSBvZgorICAgICAgICB0aGUgZm9sbG93aW5nIHJlYXNvbi4KKworICAgICAgICAxLiBBY2NvcmRp bmcgdG8gdGhlIFY4LCB0aGlzIGRlZmF1bHQgdmFsdWUgaXMgY29tcGF0aWJsZSB0byB0aGUgb2xk IEtKUy4gV2hpbGUgY3VycmVudCBWOCB1c2VzIDIwMDEsCisgICAgICAgICAgIFNwaWRlck1vbmtl eSBkb2VzIG5vdCBoYXZlIHN1Y2ggYSBmYWxsYmFjayBwYXRoLiBTbyB0aGlzIHJlYXNvbiBpcyBu b3QgYmlnIGRlYWwuCisKKyAgICAgICAgMi4gSXQgaXMgYSBsZWFwIHllYXIuIFdoZW4gdXNpbmcg YG5ldyBEYXRlKCJGZWIgMjkiKWAsIHdlIGFzc3VtZSB0aGF0IHBlb3BsZSB3YW50IHRvIHNhdmUg bW9udGggYW5kIGRheS4KKyAgICAgICAgICAgTGVhcCB5ZWFyIGNhbiBzYXZlIHVzZXIgaW5wdXRz IGlmIHRoZXkgaXMgdmFsaWQuIElmIHdlIHVzZSB0aGUgY3VycmVudCB5ZWFyIGluc3RlYWQsIHRo ZSBjdXJyZW50IHllYXIKKyAgICAgICAgICAgbWF5IG5vdCBiZSBhIGxlYXAgeWVhci4gSW4gdGhh dCBjYXNlLCBgbmV3IERhdGUoIkZlYiAyOSIpLmdldE1vbnRoKClgIGJlY29tZXMgMiAoTWFyY2gp LgorCisgICAgICAgICogd3RmL0RhdGVNYXRoLmNwcDoKKyAgICAgICAgKFdURjo6cGFyc2VEYXRl RnJvbU51bGxUZXJtaW5hdGVkQ2hhcmFjdGVycyk6CisKIDIwMTctMDQtMTIgIFl1c3VrZSBTdXp1 a2kgIDx1dGF0YW5lLnRlYUBnbWFpbC5jb20+CiAKICAgICAgICAgVXNlIEhBVkUoTUFDSElORV9D T05URVhUKSBpbnN0ZWFkIG9mIFVTRShNQUNISU5FX0NPTlRFWFQpCmRpZmYgLS1naXQgYS9Tb3Vy Y2UvV1RGL3d0Zi9EYXRlTWF0aC5jcHAgYi9Tb3VyY2UvV1RGL3d0Zi9EYXRlTWF0aC5jcHAKaW5k ZXggMTM1MGUzMDQwNjQ3OTAzZTVlMzg4YWY2ZjVmNWQ5YWU0ZTBhMWY2Yi4uMWVmMTRmOTY0N2M4 NDMyMTcwYjQzZWE4MTM1YWY5NGIxZWYxZWE4MCAxMDA2NDQKLS0tIGEvU291cmNlL1dURi93dGYv RGF0ZU1hdGguY3BwCisrKyBiL1NvdXJjZS9XVEYvd3RmL0RhdGVNYXRoLmNwcApAQCAtODkxLDEz ICs4OTEsMTAgQEAgZG91YmxlIHBhcnNlRGF0ZUZyb21OdWxsVGVybWluYXRlZENoYXJhY3RlcnMo Y29uc3QgY2hhciogZGF0ZVN0cmluZywgYm9vbCYgaGF2ZVQKICAgICAgICAgcmV0dXJuIHN0ZDo6 bnVtZXJpY19saW1pdHM8ZG91YmxlPjo6cXVpZXRfTmFOKCk7CiAgICAgZGF0ZVN0cmluZyA9IG5l d1Bvc1N0cjsKIAotICAgIGlmICghKmRhdGVTdHJpbmcpCi0gICAgICAgIHJldHVybiBzdGQ6Om51 bWVyaWNfbGltaXRzPGRvdWJsZT46OnF1aWV0X05hTigpOwotCiAgICAgaWYgKGRheSA8IDApCiAg ICAgICAgIHJldHVybiBzdGQ6Om51bWVyaWNfbGltaXRzPGRvdWJsZT46OnF1aWV0X05hTigpOwog Ci0gICAgaW50IHllYXIgPSAwOworICAgIHN0ZDo6b3B0aW9uYWw8aW50PiB5ZWFyID0gc3RkOjpu dWxsb3B0OwogICAgIGlmIChkYXkgPiAzMSkgewogICAgICAgICAvLyAjIyMgd2hlcmUgaXMgdGhl IGJvdW5kYXJ5IGFuZCB3aGF0IGhhcHBlbnMgYmVsb3c/CiAgICAgICAgIGlmICgqZGF0ZVN0cmlu ZyAhPSAnLycpCkBAIC05NjEsOSArOTU4LDExIEBAIGRvdWJsZSBwYXJzZURhdGVGcm9tTnVsbFRl cm1pbmF0ZWRDaGFyYWN0ZXJzKGNvbnN0IGNoYXIqIGRhdGVTdHJpbmcsIGJvb2wmIGhhdmVUCiAg ICAgICAgIHJldHVybiBzdGQ6Om51bWVyaWNfbGltaXRzPGRvdWJsZT46OnF1aWV0X05hTigpOwog CiAgICAgLy8gJzk5IDIzOjEyOjQwIEdNVCcKLSAgICBpZiAoeWVhciA8PSAwICYmICpkYXRlU3Ry aW5nKSB7Ci0gICAgICAgIGlmICghcGFyc2VJbnQoZGF0ZVN0cmluZywgJm5ld1Bvc1N0ciwgMTAs ICZ5ZWFyKSkKKyAgICBpZiAoKmRhdGVTdHJpbmcgJiYgIXllYXIpIHsKKyAgICAgICAgaW50IHJl c3VsdCA9IDA7CisgICAgICAgIGlmICghcGFyc2VJbnQoZGF0ZVN0cmluZywgJm5ld1Bvc1N0ciwg MTAsICZyZXN1bHQpKQogICAgICAgICAgICAgcmV0dXJuIHN0ZDo6bnVtZXJpY19saW1pdHM8ZG91 YmxlPjo6cXVpZXRfTmFOKCk7CisgICAgICAgIHllYXIgPSByZXN1bHQ7CiAgICAgfQogCiAgICAg Ly8gRG9uJ3QgZmFpbCBpZiB0aGUgdGltZSBpcyBtaXNzaW5nLgpAQCAtOTc4LDcgKzk3Nyw3IEBA IGRvdWJsZSBwYXJzZURhdGVGcm9tTnVsbFRlcm1pbmF0ZWRDaGFyYWN0ZXJzKGNvbnN0IGNoYXIq IGRhdGVTdHJpbmcsIGJvb2wmIGhhdmVUCiAgICAgICAgICAgICBpZiAoKm5ld1Bvc1N0ciAhPSAn OicpCiAgICAgICAgICAgICAgICAgcmV0dXJuIHN0ZDo6bnVtZXJpY19saW1pdHM8ZG91YmxlPjo6 cXVpZXRfTmFOKCk7CiAgICAgICAgICAgICAvLyBUaGVyZSB3YXMgbm8geWVhcjsgdGhlIG51bWJl ciB3YXMgdGhlIGhvdXIuCi0gICAgICAgICAgICB5ZWFyID0gLTE7CisgICAgICAgICAgICB5ZWFy ID0gc3RkOjpudWxsb3B0OwogICAgICAgICB9IGVsc2UgewogICAgICAgICAgICAgLy8gaW4gdGhl IG5vcm1hbCBjYXNlICh3ZSBwYXJzZWQgdGhlIHllYXIpLCBhZHZhbmNlIHRvIHRoZSBuZXh0IG51 bWJlcgogICAgICAgICAgICAgZGF0ZVN0cmluZyA9ICsrbmV3UG9zU3RyOwpAQCAtMTA0OCw5ICsx MDQ3LDExIEBAIGRvdWJsZSBwYXJzZURhdGVGcm9tTnVsbFRlcm1pbmF0ZWRDaGFyYWN0ZXJzKGNv bnN0IGNoYXIqIGRhdGVTdHJpbmcsIGJvb2wmIGhhdmVUCiAgICAgfQogICAgIAogICAgIC8vIFRo ZSB5ZWFyIG1heSBiZSBhZnRlciB0aGUgdGltZSBidXQgYmVmb3JlIHRoZSB0aW1lIHpvbmUuCi0g ICAgaWYgKGlzQVNDSUlEaWdpdCgqZGF0ZVN0cmluZykgJiYgeWVhciA9PSAtMSkgewotICAgICAg ICBpZiAoIXBhcnNlSW50KGRhdGVTdHJpbmcsICZuZXdQb3NTdHIsIDEwLCAmeWVhcikpCisgICAg aWYgKGlzQVNDSUlEaWdpdCgqZGF0ZVN0cmluZykgJiYgIXllYXIpIHsKKyAgICAgICAgaW50IHJl c3VsdCA9IDA7CisgICAgICAgIGlmICghcGFyc2VJbnQoZGF0ZVN0cmluZywgJm5ld1Bvc1N0ciwg MTAsICZyZXN1bHQpKQogICAgICAgICAgICAgcmV0dXJuIHN0ZDo6bnVtZXJpY19saW1pdHM8ZG91 YmxlPjo6cXVpZXRfTmFOKCk7CisgICAgICAgIHllYXIgPSByZXN1bHQ7CiAgICAgICAgIGRhdGVT dHJpbmcgPSBuZXdQb3NTdHI7CiAgICAgICAgIHNraXBTcGFjZXNBbmRDb21tZW50cyhkYXRlU3Ry aW5nKTsKICAgICB9CkBAIC0xMTAyLDkgKzExMDMsMTEgQEAgZG91YmxlIHBhcnNlRGF0ZUZyb21O dWxsVGVybWluYXRlZENoYXJhY3RlcnMoY29uc3QgY2hhciogZGF0ZVN0cmluZywgYm9vbCYgaGF2 ZVQKIAogICAgIHNraXBTcGFjZXNBbmRDb21tZW50cyhkYXRlU3RyaW5nKTsKIAotICAgIGlmICgq ZGF0ZVN0cmluZyAmJiB5ZWFyID09IC0xKSB7Ci0gICAgICAgIGlmICghcGFyc2VJbnQoZGF0ZVN0 cmluZywgJm5ld1Bvc1N0ciwgMTAsICZ5ZWFyKSkKKyAgICBpZiAoKmRhdGVTdHJpbmcgJiYgIXll YXIpIHsKKyAgICAgICAgaW50IHJlc3VsdCA9IDA7CisgICAgICAgIGlmICghcGFyc2VJbnQoZGF0 ZVN0cmluZywgJm5ld1Bvc1N0ciwgMTAsICZyZXN1bHQpKQogICAgICAgICAgICAgcmV0dXJuIHN0 ZDo6bnVtZXJpY19saW1pdHM8ZG91YmxlPjo6cXVpZXRfTmFOKCk7CisgICAgICAgIHllYXIgPSBy ZXN1bHQ7CiAgICAgICAgIGRhdGVTdHJpbmcgPSBuZXdQb3NTdHI7CiAgICAgICAgIHNraXBTcGFj ZXNBbmRDb21tZW50cyhkYXRlU3RyaW5nKTsKICAgICB9CkBAIC0xMTE0LDE0ICsxMTE3LDI3IEBA IGRvdWJsZSBwYXJzZURhdGVGcm9tTnVsbFRlcm1pbmF0ZWRDaGFyYWN0ZXJzKGNvbnN0IGNoYXIq IGRhdGVTdHJpbmcsIGJvb2wmIGhhdmVUCiAgICAgICAgIHJldHVybiBzdGQ6Om51bWVyaWNfbGlt aXRzPGRvdWJsZT46OnF1aWV0X05hTigpOwogCiAgICAgLy8gWTJLOiBIYW5kbGUgMiBkaWdpdCB5 ZWFycy4KLSAgICBpZiAoeWVhciA+PSAwICYmIHllYXIgPCAxMDApIHsKLSAgICAgICAgaWYgKHll YXIgPCA1MCkKLSAgICAgICAgICAgIHllYXIgKz0gMjAwMDsKLSAgICAgICAgZWxzZQotICAgICAg ICAgICAgeWVhciArPSAxOTAwOworICAgIGlmICh5ZWFyKSB7CisgICAgICAgIGludCB5ZWFyVmFs dWUgPSB5ZWFyLnZhbHVlKCk7CisgICAgICAgIGlmICh5ZWFyVmFsdWUgPj0gMCAmJiB5ZWFyVmFs dWUgPCAxMDApIHsKKyAgICAgICAgICAgIGlmICh5ZWFyVmFsdWUgPCA1MCkKKyAgICAgICAgICAg ICAgICB5ZWFyVmFsdWUgKz0gMjAwMDsKKyAgICAgICAgICAgIGVsc2UKKyAgICAgICAgICAgICAg ICB5ZWFyVmFsdWUgKz0gMTkwMDsKKyAgICAgICAgfQorICAgICAgICB5ZWFyID0geWVhclZhbHVl OworICAgIH0gZWxzZSB7CisgICAgICAgIC8vIFdlIHNlbGVjdCAyMDAwIGFzIGRlZmF1bHQgdmFs dWUuIFRoaXMgaXMgYmVjYXVzZSBvZiB0aGUgZm9sbG93aW5nIHJlYXNvbnMuCisgICAgICAgIC8v IDEuIEFjY29yZGluZyB0byB0aGUgVjgsIHRoaXMgZGVmYXVsdCB2YWx1ZSBpcyBjb21wYXRpYmxl IHRvIHRoZSBvbGQgS0pTLiBXaGlsZSBjdXJyZW50IFY4IHVzZXMgMjAwMSwKKyAgICAgICAgLy8g ICAgU3BpZGVyTW9ua2V5IGRvZXMgbm90IGhhdmUgc3VjaCBhIGZhbGxiYWNrIHBhdGguIFNvIHRo aXMgcmVhc29uIGlzIG5vdCBiaWcgZGVhbC4KKyAgICAgICAgLy8gMi4gSXQgaXMgYSBsZWFwIHll YXIuIFdoZW4gdXNpbmcgYG5ldyBEYXRlKCJGZWIgMjkiKWAsIHdlIGFzc3VtZSB0aGF0IHBlb3Bs ZSB3YW50IHRvIHNhdmUgbW9udGggYW5kIGRheS4KKyAgICAgICAgLy8gICAgTGVhcCB5ZWFyIGNh biBzYXZlIHVzZXIgaW5wdXRzIGlmIHRoZXkgaXMgdmFsaWQuIElmIHdlIHVzZSB0aGUgY3VycmVu dCB5ZWFyIGluc3RlYWQsIHRoZSBjdXJyZW50IHllYXIKKyAgICAgICAgLy8gICAgbWF5IG5vdCBi ZSBhIGxlYXAgeWVhci4gSW4gdGhhdCBjYXNlLCBgbmV3IERhdGUoIkZlYiAyOSIpLmdldE1vbnRo KClgIGJlY29tZXMgMiAoTWFyY2gpLgorICAgICAgICB5ZWFyID0gMjAwMDsKICAgICB9CisgICAg QVNTRVJUKHllYXIpOwogICAgIAotICAgIHJldHVybiB5bWRobXNUb1NlY29uZHMoeWVhciwgbW9u dGggKyAxLCBkYXksIGhvdXIsIG1pbnV0ZSwgc2Vjb25kKSAqIG1zUGVyU2Vjb25kOworICAgIHJl dHVybiB5bWRobXNUb1NlY29uZHMoeWVhci52YWx1ZSgpLCBtb250aCArIDEsIGRheSwgaG91ciwg bWludXRlLCBzZWNvbmQpICogbXNQZXJTZWNvbmQ7CiB9CiAKIGRvdWJsZSBwYXJzZURhdGVGcm9t TnVsbFRlcm1pbmF0ZWRDaGFyYWN0ZXJzKGNvbnN0IGNoYXIqIGRhdGVTdHJpbmcpCmRpZmYgLS1n aXQgYS9KU1Rlc3RzL0NoYW5nZUxvZyBiL0pTVGVzdHMvQ2hhbmdlTG9nCmluZGV4IDRhNjAwNjhh YTQ2N2EzMzU5OGU2OTBlNjQyYzlhZmYxMjdmNjBhMWQuLjM5NjEyMzA2ZWRhODY2YmNiM2YxM2Fi ZDIyYTg2N2E0NzA2YmZkMDIgMTAwNjQ0Ci0tLSBhL0pTVGVzdHMvQ2hhbmdlTG9nCisrKyBiL0pT VGVzdHMvQ2hhbmdlTG9nCkBAIC0xLDMgKzEsMTQgQEAKKzIwMTctMDQtMTMgIFl1c3VrZSBTdXp1 a2kgIDx1dGF0YW5lLnRlYUBnbWFpbC5jb20+CisKKyAgICAgICAgW0pTQ10gRGF0ZS5wYXJzZSBz aG91bGQgYWNjZXB0IHdpZGVyIHJhbmdlIG9mIHJlcHJlc2VudGF0aW9uCisgICAgICAgIGh0dHBz Oi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0xNzA3MjAKKworICAgICAgICBSZXZp ZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICAqIHN0cmVzcy9kYXRlLXJlbGF4ZWQu anM6IEFkZGVkLgorICAgICAgICAoc2hvdWxkQmUpOgorICAgICAgICAodGhyb3cubmV3LkVycm9y KToKKwogMjAxNy0wNC0xMiAgSm9zZXBoIFBlY29yYXJvICA8cGVjb3Jhcm9AYXBwbGUuY29tPgog CiAgICAgICAgIHRlc3QyNjI6IHRlc3QyNjIvdGVzdC9idWlsdC1pbnMvTmF0aXZlRXJyb3JzL0V2 YWxFcnJvci9wcm90by5qcwpkaWZmIC0tZ2l0IGEvSlNUZXN0cy9zdHJlc3MvZGF0ZS1yZWxheGVk LmpzIGIvSlNUZXN0cy9zdHJlc3MvZGF0ZS1yZWxheGVkLmpzCm5ldyBmaWxlIG1vZGUgMTAwNjQ0 CmluZGV4IDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAwMDAuLjA2NDFkMTNm YTE2MWQ1YWQzMmM3NjZjMDU2OGUyNWQ5OWIwNzc3N2UKLS0tIC9kZXYvbnVsbAorKysgYi9KU1Rl c3RzL3N0cmVzcy9kYXRlLXJlbGF4ZWQuanMKQEAgLTAsMCArMSw2MSBAQAorZnVuY3Rpb24gc2hv dWxkQmUoYWN0dWFsLCBleHBlY3RlZCkKK3sKKyAgICBpZiAoYWN0dWFsICE9PSBleHBlY3RlZCkK KyAgICAgICAgdGhyb3cgbmV3IEVycm9yKGBiYWQgdmFsdWU6ICR7YWN0dWFsfWApOworfQorCit7 CisgICAgbGV0IGRhdGUgPSBuZXcgRGF0ZSgiTWF5IDgiKTsKKyAgICBzaG91bGRCZShkYXRlLmdl dEZ1bGxZZWFyKCksIDIwMDApOworICAgIHNob3VsZEJlKGRhdGUuZ2V0TW9udGgoKSwgNCk7Cisg ICAgc2hvdWxkQmUoZGF0ZS5nZXREYXRlKCksIDgpOworfQoreworICAgIGxldCBkYXRlID0gbmV3 IERhdGUoIkZlYiAyOSIpOworICAgIHNob3VsZEJlKGRhdGUuZ2V0RnVsbFllYXIoKSwgMjAwMCk7 CisgICAgc2hvdWxkQmUoZGF0ZS5nZXRNb250aCgpLCAxKTsKKyAgICBzaG91bGRCZShkYXRlLmdl dERhdGUoKSwgMjkpOworfQoreworICAgIGxldCBkYXRlID0gbmV3IERhdGUoIiBNYXkgOCAiKTsK KyAgICBzaG91bGRCZShkYXRlLmdldEZ1bGxZZWFyKCksIDIwMDApOworICAgIHNob3VsZEJlKGRh dGUuZ2V0TW9udGgoKSwgNCk7CisgICAgc2hvdWxkQmUoZGF0ZS5nZXREYXRlKCksIDgpOworfQor eworICAgIGxldCBkYXRlID0gbmV3IERhdGUoIiBGZWIgMjkgIik7CisgICAgc2hvdWxkQmUoZGF0 ZS5nZXRGdWxsWWVhcigpLCAyMDAwKTsKKyAgICBzaG91bGRCZShkYXRlLmdldE1vbnRoKCksIDEp OworICAgIHNob3VsZEJlKGRhdGUuZ2V0RGF0ZSgpLCAyOSk7Cit9Cit7CisgICAgbGV0IGRhdGUg PSBuZXcgRGF0ZSgiTWF5LzgiKTsKKyAgICBzaG91bGRCZShkYXRlLmdldEZ1bGxZZWFyKCksIDIw MDApOworICAgIHNob3VsZEJlKGRhdGUuZ2V0TW9udGgoKSwgNCk7CisgICAgc2hvdWxkQmUoZGF0 ZS5nZXREYXRlKCksIDgpOworfQoreworICAgIGxldCBkYXRlID0gbmV3IERhdGUoIkZlYi8yOSIp OworICAgIHNob3VsZEJlKGRhdGUuZ2V0RnVsbFllYXIoKSwgMjAwMCk7CisgICAgc2hvdWxkQmUo ZGF0ZS5nZXRNb250aCgpLCAxKTsKKyAgICBzaG91bGRCZShkYXRlLmdldERhdGUoKSwgMjkpOwor fQoreworICAgIGxldCBkYXRlID0gbmV3IERhdGUoIk1heTgiKTsKKyAgICBzaG91bGRCZShkYXRl LmdldEZ1bGxZZWFyKCksIDIwMDApOworICAgIHNob3VsZEJlKGRhdGUuZ2V0TW9udGgoKSwgNCk7 CisgICAgc2hvdWxkQmUoZGF0ZS5nZXREYXRlKCksIDgpOworfQoreworICAgIGxldCBkYXRlID0g bmV3IERhdGUoIkZlYjI5Iik7CisgICAgc2hvdWxkQmUoZGF0ZS5nZXRGdWxsWWVhcigpLCAyMDAw KTsKKyAgICBzaG91bGRCZShkYXRlLmdldE1vbnRoKCksIDEpOworICAgIHNob3VsZEJlKGRhdGUu Z2V0RGF0ZSgpLCAyOSk7Cit9CisKK3sKKyAgICBsZXQgZGF0ZSA9IG5ldyBEYXRlKCJNYXkgOCAt MSIpOworICAgIHNob3VsZEJlKGRhdGUuZ2V0RnVsbFllYXIoKSwgLTEpOworICAgIHNob3VsZEJl KGRhdGUuZ2V0TW9udGgoKSwgNCk7CisgICAgc2hvdWxkQmUoZGF0ZS5nZXREYXRlKCksIDgpOwor fQo=