1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/intl/uconv/ucvlatin/nsUTF16ToUnicode.cpp Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,350 @@
1.4 +/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
1.5 +/* This Source Code Form is subject to the terms of the Mozilla Public
1.6 + * License, v. 2.0. If a copy of the MPL was not distributed with this
1.7 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
1.8 +
1.9 +#include "nsUTF16ToUnicode.h"
1.10 +#include "nsCharTraits.h"
1.11 +#include "mozilla/Endian.h"
1.12 +
1.13 +enum {
1.14 + STATE_NORMAL = 0,
1.15 + STATE_HALF_CODE_POINT = 1,
1.16 + STATE_FIRST_CALL = 2,
1.17 + STATE_SECOND_BYTE = STATE_FIRST_CALL | STATE_HALF_CODE_POINT,
1.18 + STATE_ODD_SURROGATE_PAIR = 4
1.19 +};
1.20 +
1.21 +nsresult
1.22 +nsUTF16ToUnicodeBase::UTF16ConvertToUnicode(const char * aSrc,
1.23 + int32_t * aSrcLength,
1.24 + char16_t * aDest,
1.25 + int32_t * aDestLength,
1.26 + bool aSwapBytes)
1.27 +{
1.28 + const char* src = aSrc;
1.29 + const char* srcEnd = aSrc + *aSrcLength;
1.30 + char16_t* dest = aDest;
1.31 + char16_t* destEnd = aDest + *aDestLength;
1.32 + char16_t oddHighSurrogate;
1.33 +
1.34 + switch(mState) {
1.35 + case STATE_FIRST_CALL:
1.36 + NS_ASSERTION(*aSrcLength > 1, "buffer too short");
1.37 + src+=2;
1.38 + mState = STATE_NORMAL;
1.39 + break;
1.40 +
1.41 + case STATE_SECOND_BYTE:
1.42 + NS_ASSERTION(*aSrcLength > 0, "buffer too short");
1.43 + src++;
1.44 + mState = STATE_NORMAL;
1.45 + break;
1.46 +
1.47 + case STATE_ODD_SURROGATE_PAIR:
1.48 + if (*aDestLength < 2)
1.49 + goto error;
1.50 + else {
1.51 + *dest++ = mOddHighSurrogate;
1.52 + *dest++ = mOddLowSurrogate;
1.53 + mOddHighSurrogate = mOddLowSurrogate = 0;
1.54 + mState = STATE_NORMAL;
1.55 + }
1.56 + break;
1.57 +
1.58 + case STATE_NORMAL:
1.59 + case STATE_HALF_CODE_POINT:
1.60 + default:
1.61 + break;
1.62 + }
1.63 +
1.64 + oddHighSurrogate = mOddHighSurrogate;
1.65 +
1.66 + if (src == srcEnd) {
1.67 + *aDestLength = dest - aDest;
1.68 + return (mState != STATE_NORMAL || oddHighSurrogate) ?
1.69 + NS_OK_UDEC_MOREINPUT : NS_OK;
1.70 + }
1.71 +
1.72 + const char* srcEvenEnd;
1.73 +
1.74 + char16_t u;
1.75 + if (mState == STATE_HALF_CODE_POINT) {
1.76 + if (dest == destEnd)
1.77 + goto error;
1.78 +
1.79 + // the 1st byte of a 16-bit code unit was stored in |mOddByte| in the
1.80 + // previous run while the 2nd byte has to come from |*src|.
1.81 + mState = STATE_NORMAL;
1.82 +#if MOZ_BIG_ENDIAN
1.83 + u = (mOddByte << 8) | uint8_t(*src++); // safe, we know we have at least one byte.
1.84 +#else
1.85 + u = (*src++ << 8) | mOddByte; // safe, we know we have at least one byte.
1.86 +#endif
1.87 + srcEvenEnd = src + ((srcEnd - src) & ~1); // handle even number of bytes in main loop
1.88 + goto have_codepoint;
1.89 + } else {
1.90 + srcEvenEnd = src + ((srcEnd - src) & ~1); // handle even number of bytes in main loop
1.91 + }
1.92 +
1.93 + while (src != srcEvenEnd) {
1.94 + if (dest == destEnd)
1.95 + goto error;
1.96 +
1.97 +#if !defined(__sparc__) && !defined(__arm__)
1.98 + u = *(const char16_t*)src;
1.99 +#else
1.100 + memcpy(&u, src, 2);
1.101 +#endif
1.102 + src += 2;
1.103 +
1.104 +have_codepoint:
1.105 + if (aSwapBytes)
1.106 + u = u << 8 | u >> 8;
1.107 +
1.108 + if (!IS_SURROGATE(u)) {
1.109 + if (oddHighSurrogate) {
1.110 + if (mErrBehavior == kOnError_Signal) {
1.111 + goto error2;
1.112 + }
1.113 + *dest++ = UCS2_REPLACEMENT_CHAR;
1.114 + if (dest == destEnd)
1.115 + goto error;
1.116 + oddHighSurrogate = 0;
1.117 + }
1.118 + *dest++ = u;
1.119 + } else if (NS_IS_HIGH_SURROGATE(u)) {
1.120 + if (oddHighSurrogate) {
1.121 + if (mErrBehavior == kOnError_Signal) {
1.122 + goto error2;
1.123 + }
1.124 + *dest++ = UCS2_REPLACEMENT_CHAR;
1.125 + if (dest == destEnd)
1.126 + goto error;
1.127 + }
1.128 + oddHighSurrogate = u;
1.129 + }
1.130 + else /* if (NS_IS_LOW_SURROGATE(u)) */ {
1.131 + if (oddHighSurrogate && *aDestLength > 1) {
1.132 + if (dest + 1 >= destEnd) {
1.133 + mOddLowSurrogate = u;
1.134 + mOddHighSurrogate = oddHighSurrogate;
1.135 + mState = STATE_ODD_SURROGATE_PAIR;
1.136 + goto error;
1.137 + }
1.138 + *dest++ = oddHighSurrogate;
1.139 + *dest++ = u;
1.140 + } else {
1.141 + if (mErrBehavior == kOnError_Signal) {
1.142 + goto error2;
1.143 + }
1.144 + *dest++ = UCS2_REPLACEMENT_CHAR;
1.145 + }
1.146 + oddHighSurrogate = 0;
1.147 + }
1.148 + }
1.149 + if (src != srcEnd) {
1.150 + // store the lead byte of a 16-bit unit for the next run.
1.151 + mOddByte = *src++;
1.152 + mState = STATE_HALF_CODE_POINT;
1.153 + }
1.154 +
1.155 + mOddHighSurrogate = oddHighSurrogate;
1.156 +
1.157 + *aDestLength = dest - aDest;
1.158 + *aSrcLength = src - aSrc;
1.159 + return (mState != STATE_NORMAL || oddHighSurrogate) ?
1.160 + NS_OK_UDEC_MOREINPUT : NS_OK;
1.161 +
1.162 +error:
1.163 + *aDestLength = dest - aDest;
1.164 + *aSrcLength = src - aSrc;
1.165 + return NS_OK_UDEC_MOREOUTPUT;
1.166 +
1.167 +error2:
1.168 + *aDestLength = dest - aDest;
1.169 + *aSrcLength = --src - aSrc;
1.170 + return NS_ERROR_ILLEGAL_INPUT;
1.171 +}
1.172 +
1.173 +NS_IMETHODIMP
1.174 +nsUTF16ToUnicodeBase::Reset()
1.175 +{
1.176 + mState = STATE_FIRST_CALL;
1.177 + mOddByte = 0;
1.178 + mOddHighSurrogate = 0;
1.179 + mOddLowSurrogate = 0;
1.180 + return NS_OK;
1.181 +}
1.182 +
1.183 +NS_IMETHODIMP
1.184 +nsUTF16ToUnicodeBase::GetMaxLength(const char * aSrc, int32_t aSrcLength,
1.185 + int32_t * aDestLength)
1.186 +{
1.187 + // the left-over data of the previous run have to be taken into account.
1.188 + *aDestLength = (aSrcLength + ((STATE_HALF_CODE_POINT & mState) ? 1 : 0)) / 2;
1.189 + if (mOddHighSurrogate)
1.190 + (*aDestLength)++;
1.191 + if (mOddLowSurrogate)
1.192 + (*aDestLength)++;
1.193 + return NS_OK;
1.194 +}
1.195 +
1.196 +
1.197 +NS_IMETHODIMP
1.198 +nsUTF16BEToUnicode::Convert(const char * aSrc, int32_t * aSrcLength,
1.199 + char16_t * aDest, int32_t * aDestLength)
1.200 +{
1.201 + switch (mState) {
1.202 + case STATE_FIRST_CALL:
1.203 + if (*aSrcLength < 2) {
1.204 + if (*aSrcLength < 1) {
1.205 + *aDestLength = 0;
1.206 + return NS_OK;
1.207 + }
1.208 + if (uint8_t(*aSrc) != 0xFE) {
1.209 + mState = STATE_NORMAL;
1.210 + break;
1.211 + }
1.212 + *aDestLength = 0;
1.213 + mState = STATE_SECOND_BYTE;
1.214 + return NS_OK_UDEC_MOREINPUT;
1.215 + }
1.216 +#if MOZ_LITTLE_ENDIAN
1.217 + // on LE machines, BE BOM is 0xFFFE
1.218 + if (0xFFFE != *((char16_t*)aSrc)) {
1.219 + mState = STATE_NORMAL;
1.220 + }
1.221 +#else
1.222 + if (0xFEFF != *((char16_t*)aSrc)) {
1.223 + mState = STATE_NORMAL;
1.224 + }
1.225 +#endif
1.226 + break;
1.227 +
1.228 + case STATE_SECOND_BYTE:
1.229 + if (*aSrcLength < 1) {
1.230 + *aDestLength = 0;
1.231 + return NS_OK_UDEC_MOREINPUT;
1.232 + }
1.233 + if (uint8_t(*aSrc) != 0xFF) {
1.234 + mOddByte = 0xFE;
1.235 + mState = STATE_HALF_CODE_POINT;
1.236 + }
1.237 + break;
1.238 + }
1.239 +
1.240 + return UTF16ConvertToUnicode(aSrc, aSrcLength, aDest, aDestLength,
1.241 + bool(MOZ_LITTLE_ENDIAN));
1.242 +}
1.243 +
1.244 +NS_IMETHODIMP
1.245 +nsUTF16LEToUnicode::Convert(const char * aSrc, int32_t * aSrcLength,
1.246 + char16_t * aDest, int32_t * aDestLength)
1.247 +{
1.248 + switch (mState) {
1.249 + case STATE_FIRST_CALL:
1.250 + if (*aSrcLength < 2) {
1.251 + if (*aSrcLength < 1) {
1.252 + *aDestLength = 0;
1.253 + return NS_OK;
1.254 + }
1.255 + if (uint8_t(*aSrc) != 0xFF) {
1.256 + mState = STATE_NORMAL;
1.257 + break;
1.258 + }
1.259 + *aDestLength = 0;
1.260 + mState = STATE_SECOND_BYTE;
1.261 + return NS_OK_UDEC_MOREINPUT;
1.262 + }
1.263 +#if MOZ_BIG_ENDIAN
1.264 + // on BE machines, LE BOM is 0xFFFE
1.265 + if (0xFFFE != *((char16_t*)aSrc)) {
1.266 + mState = STATE_NORMAL;
1.267 + }
1.268 +#else
1.269 + if (0xFEFF != *((char16_t*)aSrc)) {
1.270 + mState = STATE_NORMAL;
1.271 + }
1.272 +#endif
1.273 + break;
1.274 +
1.275 + case STATE_SECOND_BYTE:
1.276 + if (*aSrcLength < 1) {
1.277 + *aDestLength = 0;
1.278 + return NS_OK_UDEC_MOREINPUT;
1.279 + }
1.280 + if (uint8_t(*aSrc) != 0xFE) {
1.281 + mOddByte = 0xFF;
1.282 + mState = STATE_HALF_CODE_POINT;
1.283 + }
1.284 + break;
1.285 + }
1.286 +
1.287 + return UTF16ConvertToUnicode(aSrc, aSrcLength, aDest, aDestLength,
1.288 + bool(MOZ_BIG_ENDIAN));
1.289 +}
1.290 +
1.291 +NS_IMETHODIMP
1.292 +nsUTF16ToUnicode::Reset()
1.293 +{
1.294 + mEndian = kUnknown;
1.295 + mFoundBOM = false;
1.296 + return nsUTF16ToUnicodeBase::Reset();
1.297 +}
1.298 +
1.299 +NS_IMETHODIMP
1.300 +nsUTF16ToUnicode::Convert(const char * aSrc, int32_t * aSrcLength,
1.301 + char16_t * aDest, int32_t * aDestLength)
1.302 +{
1.303 + if(STATE_FIRST_CALL == mState && *aSrcLength < 2)
1.304 + {
1.305 + nsresult res = (*aSrcLength == 0) ? NS_OK : NS_ERROR_ILLEGAL_INPUT;
1.306 + *aSrcLength=0;
1.307 + *aDestLength=0;
1.308 + return res;
1.309 + }
1.310 + if(STATE_FIRST_CALL == mState) // first time called
1.311 + {
1.312 + // check if BOM (0xFEFF) is at the beginning, remove it if found, and
1.313 + // set mEndian accordingly.
1.314 + if(0xFF == uint8_t(aSrc[0]) && 0xFE == uint8_t(aSrc[1])) {
1.315 + mEndian = kLittleEndian;
1.316 + mFoundBOM = true;
1.317 + }
1.318 + else if(0xFE == uint8_t(aSrc[0]) && 0xFF == uint8_t(aSrc[1])) {
1.319 + mEndian = kBigEndian;
1.320 + mFoundBOM = true;
1.321 + }
1.322 + // BOM is not found, but we can use a simple heuristic to determine
1.323 + // the endianness. Assume the first character is [U+0001, U+00FF].
1.324 + // Not always valid, but it's very likely to hold for html/xml/css.
1.325 + else if(!aSrc[0] && aSrc[1]) { // 0x00 0xhh (hh != 00)
1.326 + mState = STATE_NORMAL;
1.327 + mEndian = kBigEndian;
1.328 + }
1.329 + else if(aSrc[0] && !aSrc[1]) { // 0xhh 0x00 (hh != 00)
1.330 + mState = STATE_NORMAL;
1.331 + mEndian = kLittleEndian;
1.332 + }
1.333 + else { // Neither BOM nor 'plausible' byte patterns at the beginning.
1.334 + // Just assume it's BE (following Unicode standard)
1.335 + // and let the garbage show up in the browser. (security concern?)
1.336 + // (bug 246194)
1.337 + mState = STATE_NORMAL;
1.338 + mEndian = kBigEndian;
1.339 + }
1.340 + }
1.341 +
1.342 + nsresult rv = UTF16ConvertToUnicode(aSrc, aSrcLength, aDest, aDestLength,
1.343 +#if MOZ_BIG_ENDIAN
1.344 + (mEndian == kLittleEndian)
1.345 +#else
1.346 + (mEndian == kBigEndian)
1.347 +#endif
1.348 + );
1.349 +
1.350 + // If BOM is not found and we're to return NS_OK, signal that BOM
1.351 + // is not found. Otherwise, return |rv| from |UTF16ConvertToUnicode|
1.352 + return (rv == NS_OK && !mFoundBOM) ? NS_OK_UDEC_NOBOMFOUND : rv;
1.353 +}
