15610 2012-01-18 17:33:48 +0000 Definition for rotate3d() doesn't match implementations 2012-02-14 17:13:38 +0000 1 1 1 Unclassified CSS Transforms unspecified All All CLOSED FIXED P2 normal --- 1 ayg dschulze cmarrin dino dschulze eoconnor smfr dave.null oldest_to_newest 62863 0 ayg 2012-01-18 17:33:48 +0000 """ This function is equivalent to matrix3d(1 + (1-cos(angle))*(x*x-1), -z*sin(angle)+(1-cos(angle))*x*y, y*sin(angle)+(1-cos(angle))*x*z, 0, z*sin(angle)+(1-cos(angle))*x*y, 1 + (1-cos(angle))*(y*y-1), -x*sin(angle)+(1-cos(angle))*y*z, 0, -y*sin(angle)+(1-cos(angle))*x*z, x*sin(angle)+(1-cos(angle))*y*z, 1 + (1-cos(angle))*(z*z-1), 0, 0, 0, 0, 1). """ http://dev.w3.org/csswg/css3-3d-transforms/#transform-functions rotate3d(0, 0, 1, 90deg) should therefore be equivalent to matrix3d(0,-1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1) Test-case 1: data:text/html,<!doctype html> <div style="height:50px;width:50px; border-top:1px solid black; transform:rotate3d(0,0,1,90deg)"></div> Test-case 2: data:text/html,<!doctype html> <div style="height:50px;width:50px; border-top:1px solid black; transform:matrix3d( 0,-1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1)"></div> Per spec these should render the same. In Firefox 12.0a1 and Chrome 17 dev, however (with prefixes added), they render differently: the first has the border turned to the right, and the second has it turned to the left. Implementations have the angle negated with respect to the spec, which works out to the same as negating sin(angle). So the spec should say: """ This function is equivalent to matrix3d(1 + (1-cos(angle))*(x*x-1), z*sin(angle)+(1-cos(angle))*x*y, -y*sin(angle)+(1-cos(angle))*x*z, 0, -z*sin(angle)+(1-cos(angle))*x*y, 1 + (1-cos(angle))*(y*y-1), x*sin(angle)+(1-cos(angle))*y*z, 0, y*sin(angle)+(1-cos(angle))*x*z, -x*sin(angle)+(1-cos(angle))*y*z, 1 + (1-cos(angle))*(z*z-1), 0, 0, 0, 0, 1). """ where I've just replaced sin(angle) with -sin(angle) consistently. 63455 1 smfr 2012-01-31 21:40:21 +0000 This seems pretty different from what WebKit actually implements: angle /= 2.0f; double sinA = sin(angle); double cosA = cos(angle); double sinA2 = sinA * sinA; double x2 = x*x; double y2 = y*y; double z2 = z*z; mat.m_matrix[0][0] = 1.0f - 2.0f * (y2 + z2) * sinA2; mat.m_matrix[0][1] = 2.0f * (x * y * sinA2 + z * sinA * cosA); mat.m_matrix[0][2] = 2.0f * (x * z * sinA2 - y * sinA * cosA); mat.m_matrix[1][0] = 2.0f * (y * x * sinA2 - z * sinA * cosA); mat.m_matrix[1][1] = 1.0f - 2.0f * (z2 + x2) * sinA2; mat.m_matrix[1][2] = 2.0f * (y * z * sinA2 + x * sinA * cosA); mat.m_matrix[2][0] = 2.0f * (z * x * sinA2 + y * sinA * cosA); mat.m_matrix[2][1] = 2.0f * (z * y * sinA2 - x * sinA * cosA); mat.m_matrix[2][2] = 1.0f - 2.0f * (x2 + y2) * sinA2; mat.m_matrix[0][3] = mat.m_matrix[1][3] = mat.m_matrix[2][3] = 0.0f; mat.m_matrix[3][0] = mat.m_matrix[3][1] = mat.m_matrix[3][2] = 0.0f; mat.m_matrix[3][3] = 1.0f; 63538 2 ayg 2012-02-01 17:07:26 +0000 Remember your trig identities. :) E.g., for the first entry, the spec says 1 + (1-cos(angle))*(x*x-1) while WebKit says 1.0f - 2.0f * (y2 + z2) * sinA2 = 1 - 2*(y*y + z*z) * sin(angle/2)^2. sin(t)^2 = (1 - cos(2t))/2 <http://en.wikipedia.org/wiki/Half-angle_formula#Power-reduction_formula>, so this is the same as 1 - (y*y + z*z) * (1 - cos(angle)). (x, y, z) is normalized, so x*x + y*y + z*z = 1, which means y*y + z*z = 1 - x*x, and WebKit's formula is the same as 1 - (1 - x*x) * (1 - cos(angle)) which if we distribute the first minus sign across (1 - x*x), becomes 1 + (x*x - 1) * (1 - cos(angle)) which is the same as the spec. If you look at the terms involving sin(), you can see that WebKit does do the opposite of the spec, as my black-box testing discovered and as I explained in comment #0: 2.0f * (x * y * sinA2 + z * sinA * cosA) = 2*x*y*sin(angle/2)^2 + 2*z*sin(angle/2)*cos(angle/2) = x*y*(1 - cos(angle)) + 2*z*sin(angle/2)*cos(angle/2) (power reduction in first term) = x*y*(1 - cos(angle)) + z*sin(angle) (double-angle in second term). Note that it's z*sin(angle), as I say in the bottom of comment #0, not -z*sin(angle) as the spec says. The same goes for the other entries. sin(angle) needs to be negated in every case. FWIW, it looks like Mozilla independently realized the spec didn't match WebKit when developing their 3D transform implementation, and made their implementation match WebKit instead of the spec for compat with web content: https://bugzilla.mozilla.org/show_bug.cgi?id=704468 63543 3 cmarrin 2012-02-01 18:24:06 +0000 ... > FWIW, it looks like Mozilla independently realized the spec didn't match WebKit > when developing their 3D transform implementation, and made their > implementation match WebKit instead of the spec for compat with web content: > > https://bugzilla.mozilla.org/show_bug.cgi?id=704468 So, the spec is describing the left-hand rule for rotations, which is not very common. WebKit (and now FF) implement the right-hand rule, which is the more common way of expressing the direction. So I'd say we should change the spec to match the two implementations. OBTW, I'm betting Chrome matches the other two, unless they've made some changes in Chrome specific parts of the code. So changing the spec would match all the major transform implementations. 63545 4 cmarrin 2012-02-01 18:25:14 +0000 This is all probably a fallout of our changing the spec a while back from row-major to column-major notation. Changing the spec certainly seems like the right solution 63992 5 1077 dschulze 2012-02-12 02:49:50 +0000 Created attachment 1077 rotate3d Can some one verify that this is the correct matrix that we want? It is the code that Chris applied to this bug. But the matrix on SVG Transform[1] looks different to the attached matrix as well as to the matrix on CSS Transforms. [1] http://www.w3.org/TR/2009/WD-SVG-Transforms-20090320/#threed-rotate-definition 64007 6 dschulze 2012-02-13 03:57:07 +0000 I added a matrix to the mathematical description section, that matches the current behavior of WebKit. If all agree with the math, we can close this bug now. 64034 7 ayg 2012-02-13 16:42:33 +0000 There are two mistakes in the new version: 1) alpha needs to be replaced by alpha/2. This is the "angle /= 2.0f;" line in WebKit. 2) The first matrix entry has x^2 + z^2, but it should by y^2 + z^2 like WebKit. Otherwise it's correct. 64037 8 dschulze 2012-02-13 16:47:39 +0000 Thanks a lot, I'll change that asap. I will also add a comment that alpha must be in radian for the current rotate3d matrix. This might be the difference to SVG Transforms. I will check this as well. 64103 9 dschulze 2012-02-14 17:00:03 +0000 I corrected the code with the commit 2012/02/13. Closing bug now. Feel free to reopen it if you find other issues with rotate3d. 64106 10 ayg 2012-02-14 17:13:38 +0000 Looks good to me. Thanks! 1077 2012-02-12 02:49:50 +0000 2012-02-12 02:49:50 +0000 rotate3d CodeCogsEqn.png image/png 6145 dschulze iVBORw0KGgoAAAANSUhEUgAAA64AAABYCAMAAADC+1rZAAAAOVBMVEX///8AAAAAAAAAAAAAAAAA AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACXHtMAAAAEnRSTlMAHn4+ vw5KOs8ur49ebk7vn99ugt1VAAAXZUlEQVR4Ae1di3biug4NZLinLY9p+f+PvbZjOdbTVpJS6IS1 pkkcaUvasiEE2DMcjukx/LrH9fr5eXrtqn5BCQ9uwK9lbFqlh2G8/4mPB9NahetkuNOsAF8uw3A+ l8NX3PkFJTyY9t/LWFqk9zEs1wdTGsK9VyE7GQaz2rVC4bu34zDc8rMQcSKH3Hf1yDYR9BK2wYcy MRo+Aht9S+zzIRnV3fMZr70M+DOMofksJ4ZHcbH9jP3Mcr3WydcM1+NkH8wON3JiPjwd5v1p73My RvHCGQODIiw7pgG7UXpLWFEBD0GydWITbyB3JcpLMQY19yZNKMtUdTC2crm+v5kZXo/nD/xEksyv I/XKy4oO0+NkdhQQJ8OPv8QhviKHB4+nYxCExmF3gQ2ccrq7hOUVsBCMHhc28x6yu4zyOxmDmksf zR1GmcnY7TJeYcqvWq7j8WheSF/DjZ7xDqFKBSf2Pjkvq2wxshfJfGIy4+6A/E5eeG+3IUIJDsIQ gHi23QV2g3aXQCpQOeORaQiCFBz4CEeBEcE2DwlnwjNn75QA/OaWlqM2XcynCR8NaAgBSRhSoblt HuEnAsbtM/x5y0ti1XKNi1FNKpz4E5fP3xgOPS7kWmDIDIPRkb34TmfA7ENbzgCQt7fwRBFvNbF4 YawXg0CSw94CiVv/oVECrkDjrCOUQA/GNjEEbyBXQnkexsyq7JNGzbbjdFZwz1RJjL3F1fKRb5l+ 63L9+xEivbHX0j/k9RbmJJSqTL1idg33kDoet3t4RFMaL/p2YjTCdBbYQNFPWyXgChTOdOj5jEAP xp5NhT3BG8iVUJ6HMaGW3iGj5h4IwT1TJTEWVugwHPMi+tbleoifet7ZZyn5Ffl9vJ7Cv6HMSSgV Tb1oES5HjpXZyJ4AxnG8XYbxHE6EP6dxRO+b4ApgvIWIOQjHgOiebV+BGLEUhIfdJeAKEGehveGW wQ1qhUBKCIEejJ38x8sYny+HsL3Gq5sJLLQ3owvkCijDEsZ85eR8hLykfLL1IHeln7FBAHdQlr0l kGm55nK+dblGLi5/6aw5TDeEQvPHr/F251e2aOpdhq/w9PJZX1G/fwHJeTuGi+5TeCW/xaJusb5b ZZLjDZ9x/GM4pNtPDINA9h96CxQKCsH8JeAKEGfD9Xw8X8kbeS2ERA/GjlxcIr2hcZd4uRKeCyBf yXvI7hwlIi2YEp5yphCDlJeWT8qJTzO1KRI21JzDJ0gHZTpj40OX6+kNrnzf4RogP4WEvl++hhOc ruqsp97lcIgL+mu6xZuN4JkTII9xMQeL9E56vIfnh/o9NcRLl9D39/wCCxhV3EW7vQVCrqQgGF5Q Aqqg5ix8UFXfHmmEEOkBbPA9xBuG79chNSP2reQ7XekcRXIBBRPrZcxXTo5lVoXziUe4K1B1KTIW Ms8rEbtcZoCzj7JMFWfsu5brJX3tIv2p1ta5LMfrPd9hytUGBtBrZmRtSO5fb3ED19Dn8CQ1zZJk Ef9AUQB5+Lr/OcIizWu2GOfLlAxxDxfN6QEYs92yvd4CIddw96suCIYXlAAVCJzVq3VohJjaQegB bPC95oFpG+72l3zTciXe0CBAwcz6GfOUk2OZVUlTte4KVF2KjIXM80rEhpoL3T7KMlWcse9arrgn +ehSVmuoFzEZD/BrZj4d3qmAZRqJRuHpvH6UosDwdDl/hU+AK0aLSRhNMypD3OFbxJVBDe3d7y8Q ck1VzwXBsL8EVAHmLF2rllLsECI9BTv7wgd20zZOIcjXIreglFTCzgLGPOXkUHZVdT55H02zBYyV 5Qrz3EdZpoozdpouhvP3G8JBfeEo1GEONXzf42pN10kVyglySs/J9B1WdAC2klO8DMMvwyP9bkZ8 ZR0+ciXzms0xp3iX6artnuMxjGzs3CwpkBcUgvpLwBVgztJNIcqsEkKiB2OH9N7zLYZ0RRjexJ4A TPIesjtDieQuYcxTTowRHlJeYj6TebhiY9NMbYqEDTUDnJMyg7H0mgYL4FuX6/sx3lqjy3VIt5rC fYtLWLepE3ONaQ9PvfBW9IRfhsdwNYkex3iZfY4/Vgjb+Q/YpHiHtFxvf8dTuvXFMMDYtV1UIC8o xPSXgCtAnL2fL+PIvk6mhRDowdiRks/Yxdth+EyfAl5KvlMzZXI5SpjEC6aEr5yYbXh0VTWZpr9S VxyMDbxYD2XZm4OEmRFvzMTbrfGxarmOx4/7+chv7U7Q4SPX+MknvJ7BYHodDH37HG/X24heSLMJ mnrhSw63I759nFic4cKHqAEnII0hm8P8p1iE193weA9TeDxdjtOtJoZRrD07iwrkBYWQ/hJwBTVn p/Pw/veLvriqIQR6MHZi5Bg+yIk8hu0xQOd8czNlcgWUJVPCWU7uX19V2ThspK5oTRGwp+evGS7u OSjLVEmMDedxuMLdoFXLFWfXfXSDu0iKRz31ksn8Ti8dvqXXR8VZGJbieTEEWH1ICoisSUHonHgg IOIKGGcijDjYxBa9YFDwHnJqOENwELcSimjYPSggtvLp7oqADTX35Ce4m4wdxjLhf2K5DtPbSLWy 9FE6nI3fwYJLgWnsUH8GC2bmlsfzY5gB6EkecLbgBc3n9D2GSCpAnOkw4pkWtuhUBpn3kFMjGRYH cYejiGaOQYZo5uPrCsOGmvvyY+7djP3Ich3htb2juo/TcMbm5/Jc0+GeTHg8P0ZvLCXg7M4Lms/p e6yEDStYh828h5yaK0OOonPRd4Yhmvn4usKwoeZ1qZkZJugfWa78+3F6mYcb+TJdurWo24tnCMR0 e1K03GiQBqxgWUHVOWOXIC5hQUVfh028gVxnhhRFTbb7BEG083F2hWBDzb25Efecmp1hwv6Z5Zp+ 19ZbG7HT72wRw/qQOJHD2nKj/e0jYER8tDZpjIaP2tjEPh+S0SaM174JSCbZtvgEjRw2c8P2+QgP ihg/tFzFXPbBnYGdAZOBRy5Xrzaamfh+8hsYeP0OvX4FZlsfuFxBG83MZz/ZwUDvnOy1g5Cv36HX rwB6IW+7lmv1xV8ZpWsUtNGSMYbER11ohhFGw0eGWz5F7PMhGW3DbG1RJdA7J8GucrWzqjtEnMih jbPo7CYRHlYBzhYfdZSPHfIRHhRRepZr/iWN6G8Ocs09kFAjkB0acFogHmMdNvEGjbsVGWqZu8br tOo5aYGAnZE7Z+/VtCMfUQGPUbcjtMBgWGyQ7N6BUi/Xh4nUcWlCpP4gVqgN/qyun5aVMW6LR/b2 AJ73jEDp1K/XjmT9j9cU4cHUBn1ihQkD/rAYDNw3fzV3GUVRQnyYSJ2gtCdqwAFb5ratUufBFmzz kHDGzEs52RCP7O1BnpMQRNVBnOz03Cl7qoyg0DMI7tr2FtgNurgClTMemsbgfPIRjlJGuHEe4SeC j6aE+DCROkmaUNKAK/V5dnSZuR4UwdvS9euBpDbmbw47ewCakICtfWMY7HrZjb+Pil/oNmiAkAu3 nQUuRJ9+4dVXgcZZR2iBnl6GI7ruLqFoSogPE6mTpAklDbgO4rgJSMxUZxzYgrel61cF6d41l2tf D2BVlZjK1Ct2nQwUkTuDhhJz2U5fgcuwg5ejAoWzntACPZ0MJ3TdXUIJb1fDj3vy14yr966mSJ2s FBd+lpblDHGV8WeuujJh+eF9pZ43/fofo5Sjp9X1Kxk6dszlavYgdC0JHJY5WaLiqQc9me04u3aH QECgapAk9Vfi9++YBWqTTBR2XFuBzBmuRIkh0MMZDr+87hePzO4Ciin+osr6yfp98dflgpwhyOTJ yoRulTqPEJ6kUscV8LbU9cMNbh+ZyzW6qz2QBQ6jC5p6Qk8YA3aHJBIFqb8YeclDLVCZZGLdqyto cRYqU2JI9DCGQx+/Rwnxf/BkESQzNN1CrBQ3gNpbFHYJvxIEOUMYVhTkoLX5OQSr581JgFnZhu9S ljf7EGKW4ouODZU6wAZnn0gdXAwASslr2U5ruao9IIqAdXQ09XBPJjPIHRiwOyQ2CGio4y7aVwvU JhkSdtysAoOzRgyRHmC4rA3fJMvuBaUwS15d/zdUy9UQqauV4oraWwQFGZmw21CQgxSmcol6HiQq qdTNq7WEGDwqdYAN+flE6mCeAgqUsXBbLVexVqMH1dNWCS7qIFY9mQwhd2BAYS+jig0CGkrkpTtG geokK0/XZQKsqaDFWYMlkR5guCTom2TZvaAUctlyneePJVKHlOJA7S2C1jpKIOUCMnkxeoKfY8SR 9L45/3YfpAl5ohF8esSLEnhAiCLFh2PY2NnZJ1IH89TKELLr2CIquL3VA6wIWPmiV4owXvckmZXc gT6zQyKJQEMVdtGuVaA2yaQJsLYCizObJZGewjCsDd8ky+4zCnCrKyGm39tRIbRJ9k1WiouQkpwh yOTNKxXNUUlJzlKpizc4h2rFpkqUGF3Y2+n6pVR8fxAVzNXsgUhEhCBTj/WEsauwl9ORSORSfyz3 rgGzQEmOMKBKda+uoMVZiKvEkOhhDPvEIxcoIVoidWG5MkHCyKMsZ5gUonRlQqdKnU8Ir0sBzyNS Bxp3kkpd1wQlRuZyNXsgExHh66kn9YTl3uiQQCLQQKrxHpoFypNsEOteXUGLs0hr/LKgMI8FehjD PvHI7C6gqEqIlqxf/GyXChKGUhQ5Q01BDnorKMklZuA82jqF8PqwHSJ1oHGnZ4jSbRw0xCOtHihE xID11JN6wnJvdEggEWho1Nc6bRWoTDK57tUVtDgLlWgxBHoYw7Er/eKR2V1CcSohzrpt3UpxrZ7N kMWypVJXDFs767AFb9C42yzDVgXCeSmt2qyeevU47Htzl+J5MSB213YOuNEkmwHn+LiCFmezH9sT wDE280ADuruM4lNCTDeGfEpxKDvhQJOCE0zdQ+uwmTdo3Jk6eu4kvQ48LYTQ0EH0587j+TFQhq2D FHDLSdasoMGZmTAD97GjubdRqm81aQkm2TefUpwGBeNMSa6tAQeuze06bOYNGncbZtgsgRvwtLiN PuLPncfzY+j5CGdSwC0n2bdWwMB97GjubZSO5Zp0C51KcUJD0JAsBYdMFh+swybeoHHXoVK3OOEe R5pWjw/YLMmdxluCAfG7tjHgppPsWysg4F52ZPcOlJ7lSiTluthvGWHVN3zU8m2dx2j4qOWbvjpU G2V3L0oNscn+igQWuRIncrhJSRhk8wgEkBzi4O4jjIaPOsCwQz7CgyJKWK7Vt5pEk31wZ2Bn4DkY uKMvIX5vTl6lr+/N5hXQd8aevUsP7lBYrubn9hvSBUpfG0L+cqidsc0a3LmsOs1KWo/uUNd716FD o61UoO6A0lc0IIDkUIXoOkHAyGETAtvnIzzYxGAFdngIJjpj7nwE9GoIw+GjykzZJfb5kIwqvt84 XCfQuazArHY1M6w7RCZ1N4YZgJzsWq5EuI1A6IdcQW5znT0egibboS9XV0Dcs7cTZCAodQBzn5cj M+bNpw7KY5BsneDEG2QBnSh1hpvso7TQstLhwczKndEHKncoYLzNrcdZfAYt114Zvt5oTEEuPnuF B1OCW65Sx0KsxGbuWZ1OFqkLwjBn9r+UiwWmutt/WDkaY0o+7QjhP1BOX3mtLLWaKxNjl3lDM1ek iMNtox05wLLC4OwomRm5U/pyh/ikNjBYUGtAUUIcXlGlbv4h5FSxICYnDKnscNs8wk9EjJ9jTM5H ras+0abMAy7Y5iHhTJ1G7/5G2pEDLKscVxNCnMyM3Al9oF7HPfhIb8nITlNCHH6DSp2gMgdShogE 5UBwz+p0kkjdTzIm5qNUZQ/rNdt+01nBGwjfKkXzXmjvpIVlBTUpXxkGs97c42/7ovKiIW8IERdu NSXE4Teo1AkqcyBl2EOX4J7V6SSRup9kTMynp0Rmo9fMTIUBwRsI3ypFc7l2TtqyrKAEebkWs87c Z/U6gYdODMhI2Ya3q+HHPflrxvV7V1OlLrgkGT4CCqp7eFhRkMtG8JN5XWcPcSnr4ykh2tgpiU1E 6sLthPgfud/Ts2tNACQhMSZXMyjl6IxNkgZ11Grf1RXItmqHAO5gDEQTBZQqx/5dc7naLQAi5mUF YcUpNpsJuW/bIsijtSXiLyP+3FVVqRPl6BYpIUpCckRnD3Ep6uM5ROoIduRnG5G6iWknY2I1muZe 7qXEmCC9l62T+kK/PmUfuIcxINxKseTasWMu1+ivtiC8WApEpJCNKcZzV2Zczr+PxWzs2ZjLVVWp S9+lLe+yQSpukRJifuLCQohYCajmEuvjQWRFys/GBueNROoS607GcDVFNk8pJ/dVrApeFIXeYy1E qFqJYYKDr48xaKaRopC1OtRarnoLMBEogD7FJjPIHRjQpDczqMkiCuw8YMv1v7cZwVCpq2X4QCou OlaqezC8sUpdrY/XCDGxRkQWYfZsLVKXaHMzVldTMlrAGMwnUU7R0RWRMgAHun2yfkA4oMzza9Fe tVzFYs0WVNMTggtCiKgpyQ5yBwY06c0MarIIgZds8XJ9+6++GLZU6pAMH0jFhQRq1T0Y3lalDunj 2SEm1ojIIsyeoME4EbaRSF0E8zOGqoGMFF3HKV2/dKSnKyJlMFkhPx9jQHhByXUs3FTLVUIwW4Cm Z+1dv7pOdGEdi5I7TLjNW1Rno+/rSojTbzs1KcR404zpEYZr5Hv43Q/5AoeiIJdzkoTkiM4e5lLS x1NCdGCHLFxKiIZIXUCKpPgYk6rRNPcMxrj0XjZOG0dXJMoYuIsxaCZDqTN07NvL1WyBOD1T6MYU 47krMy7X0cWio+bZNL0cwjVCfWfYVKkT5egk1b0QKClECQpyOQdBSI7o7BEuBRFGLUQbOybhUUI0 ROoGP2OKpKRWjs6YJJqXrZ1dESjj4B7GoJkcJSfo3JjL1WqBQkQK35hiPPdNW+RgIN1y+Mqv8fVy tVTqZDk6SXUvZKIpyOUkBSE5orOHuZREGLUQbeypWduI1PkZi5+nC5KSWjk6Y2n25NNk4+yKQJkA 7pD1g2YKKCTTnsMV2pEKESlqY4rx3DdtUU/hYHMeh+v05d3wXE8+yAGbaisIt1VnF+xKgFgDDnMZ QuD3FUbMNrbhHK7q2WeoS4QQBRQUtbsa8BIAMWFguGS7DlzwXkLZksQtHyktZN+YYm56hYBuDJTg fOBTQhyYcNuMtGyPAxINOKRS59PHa2LbKTP3nBlJ0AaxGPNVA3G0tOD8qu06cOb9EtqR0zdcgDbW FF+7EwzjYQEG5KNuO15dw/9TqbovO8EBLQ04nz6eD5vlz9xzZlaCDMRizFcNQGtpwflV23XgzPsF tSNZU3ztTuwzHhZgNNvYs1yTFGITyWNAlOBAblCGcOrjubB5ROKe1ek6ROoQFEGpzjmrAU8C6M0H YOTtOnDiDc3cNkU5cWuUpmXZ0qYsyp0EXIRhJRnPdS3X7aUQiegbOWwlbZ8nYOTQ9g1nsX0+woNN DIrS4dAywRngo5Zv8zyGw0dNZ8IYHHpR2nGcFisSWOaKvfCRM3fNvG+5at77+M7AzsADGdiX6wPJ 3kPtDKxjYF+u6/jbvXcGHsjAvlwfSPYeamdgHQP7cl3H3+69M/BABp52udb6bw/k45VD7ZR5u/ca jFVKkM+6XJH+m7cL/6b9Tpm37y/BGFKCfNblivTfvG34N+13yrx9fxHGql8kPetyDXnN+m/eNvyb 9jtl3r6/CGPPv1yx5IW3Df+k/U6Zt+2vwti+XL2dfQH7V5l8z0PlqzC2L9fnmTObZfIqk2+zglcD vQpj+3Jd3ernA3iVyfc8zL0KY8+/XLH+2/N0+Ikz2SnzNudVGHv+5TrpTYL+m7cP/6Y9ksz7Nylw Vv0ijL3AckX6b84u/KPmO2Xexr8IYy+wXIda/83bhn/UfqfM2/hXYAwpQT7r1ySCHsMY/3u3/eFg YKfMQVYyfTXGnne5epnf7XcGfj0D+3L99S3eC/w9DOzL9ff0cq/k1zOQlmv4f6HH8h9t/fqS9wJ3 Bl6QgbRI03K9x8cLVrCnvDPwzzCQFul9/D9hXCsoFtkq4wAAAABJRU5ErkJggg==