1--TEST-- 2Exhaustive test of mUTF-7 (IMAP) encoding verification and conversion 3--EXTENSIONS-- 4mbstring 5--FILE-- 6<?php 7include('encoding_tests.inc'); 8mb_substitute_character(0x25); // '%' 9 10function utf16BE($utf8) { 11 return mb_convert_encoding($utf8, 'UTF-16BE', 'UTF-8'); 12} 13 14function mBase64($str) { 15 return str_replace('=', '', str_replace('/', ',', base64_encode($str))); 16} 17 18function testValid($from, $to, $bothWays = true) { 19 testValidString($from, $to, 'UTF7-IMAP', 'UTF-8', $bothWays); 20} 21function testInvalid($from, $to) { 22 testInvalidString($from, $to, 'UTF7-IMAP', 'UTF-8'); 23} 24 25/* An empty string is valid */ 26testValid("", ""); 27echo "Identification passes on empty string... good start!\n"; 28 29/* Identification and conversion of ASCII characters (minus &) */ 30for ($i = 0x20; $i <= 0x7E; $i++) { 31 if ($i == 0x26) // '&' 32 continue; 33 testValid(chr($i), chr($i)); 34} 35echo "Testing all valid single-character ASCII strings... check!\n"; 36 37/* Identification and conversion of non-ASCII characters */ 38for ($i = 0; $i < 0x20; $i++) 39 testInvalid(chr($i), "%"); 40for ($i = 0x7F; $i < 256; $i++) 41 testInvalid(chr($i), "%"); 42echo "Non-ASCII characters convert to illegal char marker... yes!\n"; 43 44/* Identification of '&' when Base-64 encoded */ 45testValid("&" . mBase64(utf16BE("&")) . "-", "&", false); 46echo "& can be Base64-encoded... yes!\n"; 47 48/* Identification of unterminated & section */ 49identifyInvalidString("&", 'UTF7-IMAP'); 50identifyInvalidString("abc&", 'UTF7-IMAP'); 51identifyInvalidString("&" . mBase64(utf16BE("ハムサンドイッチ")), 'UTF7-IMAP'); 52echo "Testing unterminated & sections... yep!\n"; 53 54/* Identification of null shifts (& immediately after -) 55 * 56 * This is illegal according to the spec for mUTF-7 (IMAP), but currently we are letting 57 * it pass... among other things, this makes it possible to concatenate UTF-7-IMAP 58 * strings naively without the concatenated strings being treated as 'invalid' 59 * 60 * If ever we want to enforce this part of the spec, uncomment the following test */ 61/* 62identifyInvalidString("&" . mBase64(utf16BE("肉包子")) . "-&" . mBase64(utf16BE("冰淇淋")) . "-", 'UTF7-IMAP'); 63echo "Testing consecutive & sections which should have been merged... yep!\n"; 64*/ 65 66/* Conversion of Base64-encoded ASCII characters (excluding &) 67 * These should be treated as erroneous and mb_substitute_character should apply */ 68for ($i = 0x20; $i <= 0x7E; $i++) { 69 if ($i == 0x26) // '&' 70 continue; 71 testInvalid("&" . mBase64(utf16BE(chr($i))) . "-", "%"); 72} 73echo "Testing ASCII characters which are Base64-encoded... great!\n"; 74 75/* Conversion of & encoded as &- */ 76testValid("&-", "&"); 77testValid("abc&-", "abc&"); 78testValid("&-.&-", "&.&"); 79echo "Testing valid strings which use '&-' for '&'... good!\n"; 80 81/* Identification of & sections containing non-Base64 */ 82 83/* We'll use 6 character strings as a test, since 6 UTF-16 characters is just enough 84 * to fit perfectly in Base64 encoding, with no padding */ 85$testString = mBase64(utf16BE("我是打酱油的")); 86if (strlen($testString) != 16) 87 die("Erk!!"); 88for ($i = 0; $i < 256; $i++) { 89 if ($i >= 0x30 && $i <= 0x39) // '0'..'9' 90 continue; 91 if ($i >= 0x41 && $i <= 0x5A) // 'A'..'Z' 92 continue; 93 if ($i >= 0x61 && $i <= 0x7A) // 'a'..'z' 94 continue; 95 if ($i == 0x2B || $i == 0x2C) // '+' or ',' 96 continue; 97 if ($i == 0x2D) // '-'... this will be interpreted as ending the Base64 section 98 continue; 99 identifyInvalidString("&" . substr($testString, 0, 11) . chr($i) . "-", 'UTF7-IMAP'); 100} 101echo "Identification fails when Base64 sections contain non-Base64 bytes... right!\n"; 102 103/* Tell me, please, how many ways can UTF-16BE text get messed up? 104 * Why, that's elementary... */ 105 106/* 1. The second half of a surrogate pair could come first, */ 107$testString = mb_convert_encoding("\x00\x01\x04\x00", 'UTF-16BE', 'UTF-32BE'); 108if (strlen($testString) != 4) 109 die("Ouch!"); 110$testString = substr($testString, 2, 2) . substr($testString, 0, 2); 111identifyInvalidString("&" . mBase64($testString) . "-", 'UTF7-IMAP'); 112/* (Or could appear by itself) */ 113$testString = substr($testString, 0, 2); 114identifyInvalidString("&" . mBase64($testString) . "-", 'UTF7-IMAP'); 115 116/* ...and we should detect this wherever it occurs */ 117$singleChar = mb_convert_encoding("1", 'UTF-16BE', 'ASCII'); 118$doubleChar = mb_convert_encoding("\x00\x01\x04\x01", 'UTF-16BE', 'UTF-32BE'); 119if (strlen($doubleChar) != 4) 120 die("That was supposed to be a surrogate pair"); 121identifyInvalidString("&" . mBase64($singleChar . $testString) . "-", 'UTF7-IMAP'); 122identifyInvalidString("&" . mBase64($singleChar . $singleChar . $testString) . "-", 'UTF7-IMAP'); 123identifyInvalidString("&" . mBase64($singleChar . $singleChar . $singleChar . $testString) . "-", 'UTF7-IMAP'); 124identifyInvalidString("&" . mBase64($doubleChar . $testString) . "-", 'UTF7-IMAP'); 125identifyInvalidString("&" . mBase64($singleChar . $doubleChar . $testString) . "-", 'UTF7-IMAP'); 126identifyInvalidString("&" . mBase64($singleChar . $singleChar . $doubleChar . $testString) . "-", 'UTF7-IMAP'); 127 128/* 2. The first half of a surrogate pair might be followed by an invalid 2nd part, */ 129$testString = mb_convert_encoding("\x00\x01\x04\x00", 'UTF-16BE', 'UTF-32BE'); 130$testString = substr($testString, 0, 2) . mb_convert_encoding("a", 'UTF-16BE', 'ASCII'); 131identifyInvalidString("&" . mBase64($testString) . "-", 'UTF7-IMAP'); 132 133/* ...and we should also detect that wherever it occurs... */ 134identifyInvalidString("&" . mBase64($singleChar . $testString) . "-", 'UTF7-IMAP'); 135identifyInvalidString("&" . mBase64($singleChar . $singleChar . $testString) . "-", 'UTF7-IMAP'); 136identifyInvalidString("&" . mBase64($doubleChar . $testString) . "-", 'UTF7-IMAP'); 137 138/* 3. The first half of a surrogate pair could come at the end of the string, */ 139$testString = mb_convert_encoding("\x00\x01\x04\x00", 'UTF-16BE', 'UTF-32BE'); 140identifyInvalidString("&" . mBase64(substr($testString, 0, 2)) . "-", 'UTF7-IMAP'); 141identifyInvalidString("&" . mBase64($singleChar . substr($testString, 0, 2)) . "-", 'UTF7-IMAP'); 142identifyInvalidString("&" . mBase64($singleChar . $singleChar . substr($testString, 0, 2)) . "-", 'UTF7-IMAP'); 143 144/* 4. Or, it could have an odd number of bytes in it! */ 145$testString = utf16BE("ドーナツ"); 146$testString = substr($testString, 0, strlen($testString) - 1); 147identifyInvalidString("&" . mBase64($testString) . "-", 'UTF7-IMAP'); 148 149/* And there is one bonus way to discombobulate your UTF-16BE when it is Base64-encoded... 150 * The Base64 might not decode to an integral number of bytes 151 * Or, equivalently... it might not be padded with zeroes (as the RFC requires) */ 152$testString = utf16BE("☺⛑"); 153if (strlen($testString) != 4) 154 die("No good"); 155$encoded = mBase64($testString); 156if (strlen($encoded) != 6) 157 die("Don't like that"); 158/* Mess up the padding by replacing the last Base64 character with ',', 159 * which represents 63 (a number with a 1 in the last bit) */ 160identifyInvalidString("&" . substr($encoded, 0, strlen($encoded) - 1) . ",-", 'UTF7-IMAP'); 161 162echo "Identification fails when UTF-16 text is invalid... no sweat!\n"; 163 164/* OK, let's try valid Base64-encoded text now */ 165 166/* 2-byte char */ 167testValid("&" . mBase64(utf16BE("☺")) . "-", "☺"); 168/* 2 + 2 */ 169testValid("&" . mBase64(utf16BE("饺子")) . "-", "饺子"); 170/* 2 + 2 + 2 */ 171testValid("&" . mBase64(utf16BE("123")) . "-", "123"); 172/* 2 + 2 + 2 + 2 */ 173testValid("&" . mBase64(utf16BE("ᄚᄆᄇᄈ")) . "-", "ᄚᄆᄇᄈ"); 174/* 4 */ 175$longChar1 = mb_convert_encoding("\x00\x01\x04\x01", 'UTF-16BE', 'UTF-32BE'); 176$longChar2 = mb_convert_encoding("\x00\x01\x04\x01", 'UTF-8', 'UTF-32BE'); 177testValid("&" . mBase64($longChar1) . "-", $longChar2); 178/* 2 + 4 */ 179testValid("&" . mBase64(utf16BE("饼") . $longChar1) . "-", "饼" . $longChar2); 180/* 4 + 2 */ 181testValid("&" . mBase64($longChar1 . utf16BE("饼")) . "-", $longChar2 . "饼"); 182/* 2 + 4 + 2 */ 183testValid("&" . mBase64(utf16BE("☺") . $longChar1 . utf16BE("饼")) . "-", "☺" . $longChar2 . "饼"); 184/* 2 + 2 + 4 */ 185testValid("&" . mBase64(utf16BE("西瓜") . $longChar1) . "-", "西瓜" . $longChar2); 186/* 2 + 2 + 4 + 2 */ 187testValid("&" . mBase64(utf16BE("西瓜") . $longChar1 . utf16BE("☺")) . "-", "西瓜" . $longChar2 . "☺"); 188/* 2 + 2 + 4 + 4 */ 189testValid("&" . mBase64(utf16BE("西瓜") . $longChar1 . $longChar1) . "-", "西瓜" . $longChar2 . $longChar2); 190/* 2 + 2 + 2 + 4 */ 191testValid("&" . mBase64(utf16BE("西红柿") . $longChar1) . "-", "西红柿" . $longChar2); 192 193/* Multiple sections of valid ASCII _and_ Base64-encoded text */ 194testValid("123&" . mBase64(utf16BE("123")) . "-abc&" . mBase64(utf16BE("☺")) . "-.", "123123abc☺."); 195 196/* If a & character appears right after a non-ASCII character, we must first close the Base64 197 * section and then emit &- */ 198testValidString("☺&", "&Jjo-&-", "UTF-8", "UTF7-IMAP", false); 199testValidString("西瓜&", "&iX903A-&-", "UTF-8", "UTF7-IMAP", false); 200testValidString("西红柿&", "&iX9+omf,-&-", "UTF-8", "UTF7-IMAP", false); 201 202echo "Identification and conversion of valid text is working... perfect!\n"; 203 204// Try illegal Unicode codepoint (> 0x10FFFF) 205convertInvalidString("\x00\x20\x00\x00", "%", "UCS-4BE", "UTF7-IMAP"); 206 207// Test "long" illegal character markers 208mb_substitute_character("long"); 209convertInvalidString("\x10", "%", "UTF7-IMAP", "UTF-8"); 210convertInvalidString("\x80", "%", "UTF7-IMAP", "UTF-8"); 211convertInvalidString("abc&", "abc%", "UTF7-IMAP", "UTF-8"); // The & starts a Base-64 coded section, which is OK... but there's no data in it 212convertInvalidString("&**-", "%*-", "UTF7-IMAP", "UTF-8"); // When we hit the first bad byte in a Base-64 coded section, it drops us back into the default mode, so the following characters are literal 213 214echo "Done!\n"; 215?> 216--EXPECT-- 217Identification passes on empty string... good start! 218Testing all valid single-character ASCII strings... check! 219Non-ASCII characters convert to illegal char marker... yes! 220& can be Base64-encoded... yes! 221Testing unterminated & sections... yep! 222Testing ASCII characters which are Base64-encoded... great! 223Testing valid strings which use '&-' for '&'... good! 224Identification fails when Base64 sections contain non-Base64 bytes... right! 225Identification fails when UTF-16 text is invalid... no sweat! 226Identification and conversion of valid text is working... perfect! 227Done! 228