diff options
Diffstat (limited to 'thirdparty/icu4c/common/utf_impl.cpp')
-rw-r--r-- | thirdparty/icu4c/common/utf_impl.cpp | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/thirdparty/icu4c/common/utf_impl.cpp b/thirdparty/icu4c/common/utf_impl.cpp index 9dd241a12b..a1f9c6529a 100644 --- a/thirdparty/icu4c/common/utf_impl.cpp +++ b/thirdparty/icu4c/common/utf_impl.cpp @@ -55,7 +55,7 @@ * -finish: * (BSR: Bit Scan Reverse, scans for a 1-bit, starting from the MSB) */ -extern "C" U_EXPORT const uint8_t +U_CAPI const uint8_t utf8_countTrailBytes[256]={ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, @@ -117,11 +117,11 @@ errorValue(int32_t count, int8_t strict) { * Unicode 16-bit strings that are not well-formed UTF-16, that is, they * contain unpaired surrogates. * -3: All illegal byte sequences yield U+FFFD. - * 0 Obsolete "safe" behavior of UTF8_NEXT_CHAR_SAFE(..., FALSE): + * 0 Obsolete "safe" behavior of UTF8_NEXT_CHAR_SAFE(..., false): * All illegal byte sequences yield a positive code point such that this * result code point would be encoded with the same number of bytes as * the illegal sequence. - * >0 Obsolete "strict" behavior of UTF8_NEXT_CHAR_SAFE(..., TRUE): + * >0 Obsolete "strict" behavior of UTF8_NEXT_CHAR_SAFE(..., true): * Same as the obsolete "safe" behavior, but non-characters are also treated * like illegal sequences. * @@ -214,7 +214,7 @@ utf8_appendCharSafeBody(uint8_t *s, int32_t i, int32_t length, UChar32 c, UBool } /* c>0x10ffff or not enough space, write an error value */ if(pIsError!=NULL) { - *pIsError=TRUE; + *pIsError=true; } else { length-=i; if(length>0) { |