7831 2006-03-17 10:41:03 -0800 .apply/call does not set this for eval 2008-07-20 13:30:51 -0700 1 1 1 Unclassified WebKit JavaScriptCore 420+ Mac OS X 10.4 RESOLVED FIXED HasReduction P2 Normal --- 1 tolmasky webkit-unassigned ddkilzer ggaren ian oliver zwarich oldest_to_newest 36756 0 tolmasky 2006-03-17 10:41:03 -0800 Since eval simply ignores the this argument and always opts for the calling context's this or the global object, using eval.apply(something, args) does not work. ECMA is somewhat unclear about this (10.2 says to use the calling context's this object), but I think it's pretty clear that if you explicitly ask to use another this then you should, given that that is basically the point of apply/call. Firefox handles this the same way Safari currently does (incorrectly in my opinion). 36757 1 7132 tolmasky 2006-03-17 10:42:36 -0800 Created attachment 7132 test case when this.x=9 is done, window.x is set instead of a.x, which is why "undfined" is then alerted. I believe "9" should be alerted. 36759 2 7133 tolmasky 2006-03-17 10:48:46 -0800 Created attachment 7133 Makes eval.apply/call pay attention to the this argument. What this does is make callAsFunction for eval check whether there is in fact a thisObj, which only takes place I believe if its called through apply or call. In this case, if its not undefined or null, it uses it. ContextImp now also checks for a this argument in eval if which again should only take place if we called it initially through apply or call. 36769 3 7133 mjs 2006-03-17 11:48:09 -0800 Comment on attachment 7133 Makes eval.apply/call pay attention to the this argument. The substance of the change looks good. I have some style concerns: 1) I don't think you need comments explaining what the code used to do or why it was wrong. That can go in the ChangeLog. 2) This if statement shouldn't be all on one line. + if(!thisObj||thisObj->isUndefinedOrNull()) thisObj= static_cast<JSObject *>(exec->context().thisValue()); Can thisObj ever actually truly be a null pointer as opposed to JS null? 3) Needs a ChangeLog entry. 4) Needs a test case - you can look in LayoutTests/fast/js/resources/*.js, make a JS file like that using shouldBe and then use make-js-test-wrappers to make the HTML wrapper. r- for ChangeLog and test case, but the substance of the change looks correct. 36779 4 ggaren 2006-03-17 13:49:47 -0800 No, thisObj can never be a NULL pointer. It's important to leave NULL checks out of code that works with JSValues, because they might confuse folks into thinking that it's OK to return NULL rather than jsNull(). 36780 5 ggaren 2006-03-17 13:51:58 -0800 Also, please execute run-javascriptcore-tests to make sure this patch doesn't cause any regressions in the mozilla JS tests. (And maybe it will fix a few!) 36836 6 7146 tolmasky 2006-03-18 01:25:18 -0800 Created attachment 7146 Modified Patch I ended up having to make a significantly different patch for this. The old one, while working, caused a regression in Expressions/1.11.1.js or so. The problem is that callAsFunction always gets sent a thisObj, so its not enough to check whether thisObj isn't null or undefined to see whether we are actually being called from an apply or call. So the solution I came up with was to trick it by creating an inbetween ExecState that contains the correct thisObj. I wanted to float this by you guys as whether it was an acceptible solution or not before writing a changelog/tests for it, since it does seem a bit hackish. Also, this new patch fixes js1_6/Array/regress-304828.js . 36961 7 7146 darin 2006-03-19 16:30:49 -0800 Comment on attachment 7146 Modified Patch Thanks for taking this on! There are many minor things wrong with this patch. 1) no tests -- patches that fix bugs require tests that demonstrate the bug is fixed (although I suppose the fact that this fixes on of the Mozilla tests might be an OK substitute) 2) no change log entry 3) commented-out code left in internal.cpp, should not be 4) patch that changes only whitespace in function.cpp 5) tabs in the patch (we use spaces for indenting) 6) no space between "if" and the "(" following it in the code in function_object.cpp 7) no spaces around the "==" operator 8) "ContextImp *cpy= " should be "ContextImp* cpy = ". 9) We don't declare multiple variables with a single declaration as is done with "cpy" and "ctx". 10) local variable named "cpy" -- "copy" would be better 11) local variable named "ctx" -- name that is either a word or just a single letter would be better 12) no spaces after commas in some function calls However, the main issue here is almost certainly performance. Adding an inherits check to every function call is probably not a good idea. And as to the reason that's needed, I don't see what information FunctionProtoFunc::callAsFunction has that GlobalFuncImp::callAsFunction will not have. Won't the internal.cpp change fix the bug without any change to function_object.cpp at all? If not, what's the high level description of what the function_object.cpp change does? 86463 8 ddkilzer 2008-07-20 06:49:15 -0700 This appears to be fixed (the test case no longer alerts) with WebKit nightly build r35249. 86464 9 ddkilzer 2008-07-20 06:52:22 -0700 Oops...since nothing is alerted on the nightly WebKit (expected is that "9" is alerted per Comment #1), this is a change in behavior that should probably be investigated. Reopening. 86465 10 ddkilzer 2008-07-20 06:54:28 -0700 Firefox 2.0.0.x does not show an alert, either. Firebug lists the error as: function eval must be called directly, and not by way of a function of another name 86466 11 ddkilzer 2008-07-20 07:20:33 -0700 Running a bisect of nightly builds to find when the behavior changed. Interestingly, nightly builds r27811-r28007 all crashed with a null pointer running the test case. 86467 12 ddkilzer 2008-07-20 07:34:24 -0700 The bisect-builds script reports: Fails: r30868 Works: r30885 Where "Fails" means an alert was displayed with "undefined" in the text of the alert, and "Works" means no alert was displayed. Again, the expected result per Comment #1 is that "9" is displayed. 86469 13 ddkilzer 2008-07-20 07:48:07 -0700 (In reply to comment #12) > The bisect-builds script reports: > > Fails: r30868 Works: r30885 The only interesting commit in this range is: http://trac.webkit.org/changeset/30871 86470 14 22388 ddkilzer 2008-07-20 07:52:59 -0700 Created attachment 22388 test case for exception This test case tests that an exception is thrown in the expression. This seems to jibe with the changes in r30871. I'd suggest closing this bug in light of the above information. 86495 15 oliver 2008-07-20 13:30:51 -0700 This behaviour is "correct". Clearly firefox does the eval aliasing check before it does the target object checking which results in a different failure. However this doesn't change the fact that eval can not be called in this manner. 7132 2006-03-17 10:42:36 -0800 2006-03-17 10:42:36 -0800 test case tester.html text/html 70 tolmasky PHNjcmlwdD4KYT0ge307CgpldmFsLmFwcGx5KGEsIFsidGhpcy54PSA5OyBhbGVydChhLngpOyJd KTsKCjwvc2NyaXB0Pg== 7133 2006-03-17 10:48:46 -0800 2006-03-17 11:48:09 -0800 Makes eval.apply/call pay attention to the this argument. evalapplypatch.txt text/plain 2268 tolmasky SW5kZXg6IEphdmFTY3JpcHRDb3JlL2tqcy9pbnRlcm5hbC5jcHAKPT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gSmF2 YVNjcmlwdENvcmUva2pzL2ludGVybmFsLmNwcAkocmV2aXNpb24gMTMzNTQpCisrKyBKYXZhU2Ny aXB0Q29yZS9ranMvaW50ZXJuYWwuY3BwCSh3b3JraW5nIGNvcHkpCkBAIC0yMTYsNyArMjE2LDEw IEBAIENvbnRleHRJbXA6OkNvbnRleHRJbXAoSlNPYmplY3QgKmdsb2IsIEkKICAgICAgIGlmIChf Y2FsbGluZ0NvbnRleHQpIHsKICAgICAgICAgc2NvcGUgPSBfY2FsbGluZ0NvbnRleHQtPnNjb3Bl Q2hhaW4oKTsKICAgICAgICAgdmFyaWFibGUgPSBfY2FsbGluZ0NvbnRleHQtPnZhcmlhYmxlT2Jq ZWN0KCk7Ci0gICAgICAgIHRoaXNWYWwgPSBfY2FsbGluZ0NvbnRleHQtPnRoaXNWYWx1ZSgpOwor CQkvLyB3YXM6IF9jYWxsaW5nQ29udGV4dC0+dGhpc1ZhbHVlKCk7ICBIb3dldmVyLCB3ZSBzaG91 bGQgdHJ1c3QgdGhpc1YsIGJlY2F1c2UgY2FsbEFzRnVuY3Rpb24gd2lsbCBoYW5kbGUgZWl0aGVy IHNlbmRpbmcgdXMgdGhlIAorCQkvLyBjYWxsaW5nIGNvbnRleHQncyB0aGlzT2JqZWN0LCBvciBh c3NpZ24gdXMgYSBuZXcgdGhpc09iamVjdCBpZiB3ZSB3ZXJlIGNhbGxlZCB3aXRoIGFwcGx5IG9y IGNhbGwsIGluIHdoaWNoIGNhc2Ugd2Ugd2FudCB0aGUgdXNlciBkZWZpbmVkIAorCQkvLyB0aGlz IG9iamVjdC4KKyAgICAgICAgdGhpc1ZhbCA9IHRoaXNWOy8vX2NhbGxpbmdDb250ZXh0LT50aGlz VmFsdWUoKTsKICAgICAgICAgYnJlYWs7CiAgICAgICB9IC8vIGVsc2Ugc2FtZSBhcyBHbG9iYWxD b2RlCiAgICAgY2FzZSBHbG9iYWxDb2RlOgpJbmRleDogSmF2YVNjcmlwdENvcmUva2pzL2Z1bmN0 aW9uLmNwcAo9PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09Ci0tLSBKYXZhU2NyaXB0Q29yZS9ranMvZnVuY3Rpb24uY3BwCShy ZXZpc2lvbiAxMzM1NCkKKysrIEphdmFTY3JpcHRDb3JlL2tqcy9mdW5jdGlvbi5jcHAJKHdvcmtp bmcgY29weSkKQEAgLTc1MSw3ICs3NTEsNyBAQCBzdGF0aWMgZG91YmxlIHBhcnNlRmxvYXQoY29u c3QgVVN0cmluZyAmCiAgICAgcmV0dXJuIHMudG9Eb3VibGUoIHRydWUgLyp0b2xlcmFudCovLCBm YWxzZSAvKiBOYU4gZm9yIGVtcHR5IHN0cmluZyAqLyApOwogfQogCi1KU1ZhbHVlICpHbG9iYWxG dW5jSW1wOjpjYWxsQXNGdW5jdGlvbihFeGVjU3RhdGUgKmV4ZWMsIEpTT2JqZWN0ICovKnRoaXNP YmoqLywgY29uc3QgTGlzdCAmYXJncykKK0pTVmFsdWUgKkdsb2JhbEZ1bmNJbXA6OmNhbGxBc0Z1 bmN0aW9uKEV4ZWNTdGF0ZSAqZXhlYywgSlNPYmplY3QgKnRoaXNPYmosIGNvbnN0IExpc3QgJmFy Z3MpCiB7CiAgIEpTVmFsdWUgKnJlcyA9IGpzVW5kZWZpbmVkKCk7CiAKQEAgLTgwMCwxMCArODAw LDEzIEBAIEpTVmFsdWUgKkdsb2JhbEZ1bmNJbXA6OmNhbGxBc0Z1bmN0aW9uKEUKICAgICAgICAg fQogCiAgICAgICAgIC8vIGVudGVyIGEgbmV3IGV4ZWN1dGlvbiBjb250ZXh0Ci0gICAgICAgIEpT T2JqZWN0ICp0aGlzVmFsID0gc3RhdGljX2Nhc3Q8SlNPYmplY3QgKj4oZXhlYy0+Y29udGV4dCgp LnRoaXNWYWx1ZSgpKTsKKwkJLy8gV2UgYXJlIGluIGZhY3QgaW50ZXJlc3RlZCBpbiB0aGlzT2Jq IGJlY2F1c2UgZXZhbCBtYXkgaGF2ZSBoYXZlIGJlZW4gY2FsbGVkIHRocm91Z2ggYXBwbHkgb3Ig Y2FsbCwgdGh1cyAKKwkJLy8gYWxsb3dpbmcgdGhlIHVzZXIgdG8gcHJvdmlkZSBoaXMgb3IgaGVy IG93biB0aGlzIG9iamVjdC4KKwkJaWYoIXRoaXNPYmp8fHRoaXNPYmotPmlzVW5kZWZpbmVkT3JO dWxsKCkpIHRoaXNPYmo9IHN0YXRpY19jYXN0PEpTT2JqZWN0ICo+KGV4ZWMtPmNvbnRleHQoKS50 aGlzVmFsdWUoKSk7CisKICAgICAgICAgQ29udGV4dEltcCBjdHgoZXhlYy0+ZHluYW1pY0ludGVy cHJldGVyKCktPmdsb2JhbE9iamVjdCgpLAogICAgICAgICAgICAgICAgICAgICAgICBleGVjLT5k eW5hbWljSW50ZXJwcmV0ZXIoKS0+aW1wKCksCi0gICAgICAgICAgICAgICAgICAgICAgIHRoaXNW YWwsCisgICAgICAgICAgICAgICAgICAgICAgIHRoaXNPYmosCiAgICAgICAgICAgICAgICAgICAg ICAgIHByb2dOb2RlLmdldCgpLAogICAgICAgICAgICAgICAgICAgICAgICBFdmFsQ29kZSwKICAg ICAgICAgICAgICAgICAgICAgICAgZXhlYy0+Y29udGV4dCgpLmltcCgpKTsK 7146 2006-03-18 01:25:18 -0800 2006-03-19 16:30:49 -0800 Modified Patch newevalapplycallpatch.txt text/plain 3469 tolmasky SW5kZXg6IEphdmFTY3JpcHRDb3JlL2tqcy9pbnRlcm5hbC5jcHAKPT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gSmF2 YVNjcmlwdENvcmUva2pzL2ludGVybmFsLmNwcAkocmV2aXNpb24gMTMzNjUpCisrKyBKYXZhU2Ny aXB0Q29yZS9ranMvaW50ZXJuYWwuY3BwCSh3b3JraW5nIGNvcHkpCkBAIC0yMTYsNyArMjE2LDcg QEAgQ29udGV4dEltcDo6Q29udGV4dEltcChKU09iamVjdCAqZ2xvYiwgSQogICAgICAgaWYgKF9j YWxsaW5nQ29udGV4dCkgewogICAgICAgICBzY29wZSA9IF9jYWxsaW5nQ29udGV4dC0+c2NvcGVD aGFpbigpOwogICAgICAgICB2YXJpYWJsZSA9IF9jYWxsaW5nQ29udGV4dC0+dmFyaWFibGVPYmpl Y3QoKTsKLSAgICAgICAgdGhpc1ZhbCA9IF9jYWxsaW5nQ29udGV4dC0+dGhpc1ZhbHVlKCk7Cisg ICAgICAgIHRoaXNWYWwgPSB0aGlzVjsvL19jYWxsaW5nQ29udGV4dC0+dGhpc1ZhbHVlKCk7CiAg ICAgICAgIGJyZWFrOwogICAgICAgfSAvLyBlbHNlIHNhbWUgYXMgR2xvYmFsQ29kZQogICAgIGNh c2UgR2xvYmFsQ29kZToKSW5kZXg6IEphdmFTY3JpcHRDb3JlL2tqcy9mdW5jdGlvbi5jcHAKPT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PQotLS0gSmF2YVNjcmlwdENvcmUva2pzL2Z1bmN0aW9uLmNwcAkocmV2aXNpb24gMTMz NjUpCisrKyBKYXZhU2NyaXB0Q29yZS9ranMvZnVuY3Rpb24uY3BwCSh3b3JraW5nIGNvcHkpCkBA IC04MDcsNyArODA3LDcgQEAgSlNWYWx1ZSAqR2xvYmFsRnVuY0ltcDo6Y2FsbEFzRnVuY3Rpb24o RQogICAgICAgICAgICAgICAgICAgICAgICBwcm9nTm9kZS5nZXQoKSwKICAgICAgICAgICAgICAg ICAgICAgICAgRXZhbENvZGUsCiAgICAgICAgICAgICAgICAgICAgICAgIGV4ZWMtPmNvbnRleHQo KS5pbXAoKSk7Ci0gICAgICAgIAorCiAgICAgICAgIEV4ZWNTdGF0ZSBuZXdFeGVjKGV4ZWMtPmR5 bmFtaWNJbnRlcnByZXRlcigpLCAmY3R4KTsKICAgICAgICAgbmV3RXhlYy5zZXRFeGNlcHRpb24o ZXhlYy0+ZXhjZXB0aW9uKCkpOyAvLyBjb3VsZCBiZSBudWxsCiAgICAgICAgIApJbmRleDogSmF2 YVNjcmlwdENvcmUva2pzL2Z1bmN0aW9uX29iamVjdC5jcHAKPT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gSmF2YVNj cmlwdENvcmUva2pzL2Z1bmN0aW9uX29iamVjdC5jcHAJKHJldmlzaW9uIDEzMzY1KQorKysgSmF2 YVNjcmlwdENvcmUva2pzL2Z1bmN0aW9uX29iamVjdC5jcHAJKHdvcmtpbmcgY29weSkKQEAgLTI5 LDYgKzI5LDcgQEAKICNpbmNsdWRlICJsZXhlci5oIgogI2luY2x1ZGUgImRlYnVnZ2VyLmgiCiAj aW5jbHVkZSAib2JqZWN0LmgiCisjaW5jbHVkZSAiY29udGV4dC5oIgogCiAjaW5jbHVkZSA8YXNz ZXJ0Lmg+CiAjaW5jbHVkZSA8c3RkaW8uaD4KQEAgLTExOSw4ICsxMjAsMjUgQEAgSlNWYWx1ZSAq RnVuY3Rpb25Qcm90b0Z1bmM6OmNhbGxBc0Z1bmN0aQogICAgICAgZWxzZQogICAgICAgICByZXR1 cm4gdGhyb3dFcnJvcihleGVjLCBUeXBlRXJyb3IpOwogICAgIH0KLSAgICByZXN1bHQgPSBmdW5j LT5jYWxsKGV4ZWMsYXBwbHlUaGlzLGFwcGx5QXJncyk7CisJCisJaWYoZnVuYy0+aW5oZXJpdHMo Jkdsb2JhbEZ1bmNJbXA6OmluZm8pICYmIHN0YXRpY19jYXN0PEdsb2JhbEZ1bmNJbXAqPihmdW5j KS0+Y29kZVR5cGUoKT09RXZhbENvZGUpIHsJCQorCQlDb250ZXh0SW1wICpjcHk9IGV4ZWMtPmNv bnRleHQoKS5pbXAoKSwKKwkJCQkJY3R4KGV4ZWMtPmR5bmFtaWNJbnRlcnByZXRlcigpLT5nbG9i YWxPYmplY3QoKSwKKwkJCQkJCWV4ZWMtPmR5bmFtaWNJbnRlcnByZXRlcigpLT5pbXAoKSwKKwkJ CQkJCWFwcGx5VGhpcywKKwkJCQkJCWNweS0+Y3VycmVudEJvZHkoKSwKKwkJCQkJCUV2YWxDb2Rl LAorCQkJCQkJY3B5KTsKKwkJICAgICAgIAorCQlFeGVjU3RhdGUgbmV3RXhlYyhleGVjLT5keW5h bWljSW50ZXJwcmV0ZXIoKSwgJmN0eCk7CisKKwkJcmVzdWx0ID0gZnVuYy0+Y2FsbCgmbmV3RXhl YywgYXBwbHlUaGlzLCBhcHBseUFyZ3MpOworCX0KKwllbHNlIHsKKwkJcmVzdWx0ID0gZnVuYy0+ Y2FsbChleGVjLGFwcGx5VGhpcyxhcHBseUFyZ3MpOwogICAgIH0KKwkKKwl9CiAgICAgYnJlYWs7 CiAgIGNhc2UgQ2FsbDogewogICAgIEpTVmFsdWUgKnRoaXNBcmcgPSBhcmdzWzBdOwpAQCAtMTM1 LDcgKzE1MywyMyBAQCBKU1ZhbHVlICpGdW5jdGlvblByb3RvRnVuYzo6Y2FsbEFzRnVuY3RpCiAg ICAgZWxzZQogICAgICAgY2FsbFRoaXMgPSB0aGlzQXJnLT50b09iamVjdChleGVjKTsKIAotICAg IHJlc3VsdCA9IGZ1bmMtPmNhbGwoZXhlYyxjYWxsVGhpcyxhcmdzLmNvcHlUYWlsKCkpOworCWlm KGZ1bmMtPmluaGVyaXRzKCZHbG9iYWxGdW5jSW1wOjppbmZvKSAmJiBzdGF0aWNfY2FzdDxHbG9i YWxGdW5jSW1wKj4oZnVuYyktPmNvZGVUeXBlKCk9PUV2YWxDb2RlKSB7CQkKKwkJQ29udGV4dElt cCAqY3B5PSBleGVjLT5jb250ZXh0KCkuaW1wKCksCisJCQkJCWN0eChleGVjLT5keW5hbWljSW50 ZXJwcmV0ZXIoKS0+Z2xvYmFsT2JqZWN0KCksCisJCQkJCQlleGVjLT5keW5hbWljSW50ZXJwcmV0 ZXIoKS0+aW1wKCksCisJCQkJCQljYWxsVGhpcywKKwkJCQkJCWNweS0+Y3VycmVudEJvZHkoKSwK KwkJCQkJCUV2YWxDb2RlLAorCQkJCQkJY3B5KTsKKwkJICAgICAgIAorCQlFeGVjU3RhdGUgbmV3 RXhlYyhleGVjLT5keW5hbWljSW50ZXJwcmV0ZXIoKSwgJmN0eCk7CisKKwkJcmVzdWx0ID0gZnVu Yy0+Y2FsbCgmbmV3RXhlYyxjYWxsVGhpcyxhcmdzLmNvcHlUYWlsKCkpOworCX0KKwllbHNlIHsK KwkJcmVzdWx0ID0gZnVuYy0+Y2FsbChleGVjLGNhbGxUaGlzLGFyZ3MuY29weVRhaWwoKSk7Cisg ICAgfQorCQogICAgIH0KICAgICBicmVhazsKICAgfQpJbmRleDogSmF2YVNjcmlwdENvcmUva2pz L2ludGVycHJldGVyLmgKPT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gSmF2YVNjcmlwdENvcmUva2pzL2ludGVycHJl dGVyLmgJKHJldmlzaW9uIDEzMzY1KQorKysgSmF2YVNjcmlwdENvcmUva2pzL2ludGVycHJldGVy LmgJKHdvcmtpbmcgY29weSkKQEAgLTQzOCw2ICs0MzgsNyBAQCBuYW1lc3BhY2UgS0pTIHsKICAg ICBmcmllbmQgY2xhc3MgRnVuY3Rpb25JbXA7CiAgICAgZnJpZW5kIGNsYXNzIFJ1bnRpbWVNZXRo b2RJbXA7CiAgICAgZnJpZW5kIGNsYXNzIEdsb2JhbEZ1bmNJbXA7CisJZnJpZW5kIGNsYXNzIEZ1 bmN0aW9uUHJvdG9GdW5jOwogICBwdWJsaWM6CiAgICAgLyoqCiAgICAgICogUmV0dXJucyB0aGUg aW50ZXJwcmV0ZXIgYXNzb2NpYXRlZCB3aXRoIHRoaXMgZXhlY3V0aW9uIHN0YXRlCg== 22388 2008-07-20 07:52:59 -0700 2008-07-20 07:52:59 -0700 test case for exception bug-7831-test-exception.html text/html 125 ddkilzer PHNjcmlwdD4KYT0ge307Cgp0cnkgewpldmFsLmFwcGx5KGEsIFsidGhpcy54PSA5OyBhbGVydChh LngpOyJdKTsKfSBjYXRjaCAoZXgpIHsKICAgIGFsZXJ0KCJFeGNlcHRpb246ICIgKyBleCk7Cn0K Cjwvc2NyaXB0Pgo=