Make the detection of BOM sizes actually work

Originally committed to SVN as r4774.
This commit is contained in:
Thomas Goyne 2010-09-09 06:47:29 +00:00
parent 6f7daef630
commit 81709164b3

View file

@ -31,8 +31,6 @@
#include <libaegisub/charset_conv.h>
#include <iconv.h>
using std::min;
// Check if we can use advanced fallback capabilities added in GNU's iconv
// implementation
#if !defined(_LIBICONV_VERSION) || _LIBICONV_VERSION < 0x010A || defined(LIBICONV_PLUG)
@ -121,7 +119,13 @@ public:
iconv_t cd = iconv_open(GetRealEncodingName(targetEnc), "UTF-8");
assert(cd != iconv_invalid);
// Get BOM size (if any)
// Most (but not all) iconv implementations automatically insert a BOM
// at the beginning of text converted to UTF-8, UTF-16 and UTF-32, but
// we usually don't want this, as some of the wxString using code
// assumes there is no BOM (as the exact encoding is known externally)
// As such, when doing conversions we will strip the BOM if it exists,
// then manually add it when writing files
const char* src = "";
char *dst = buff;
size_t srcLen = 1;
@ -130,13 +134,14 @@ public:
size_t res = iconv(cd, ICONV_CONST_CAST(&src), &srcLen, &dst, &dstLen);
assert(res != iconv_failed);
assert(srcLen == 0);
src = buff;
bomSize = 0;
for (src = buff; src < dst; ++src) {
if (*src) {
bomSize = (8 - dstLen) / 2;
break;
}
if (*src) ++bomSize;
}
if (bomSize) {
// If there is a BOM, it will always be at least as big as the NUL
bomSize = std::min(bomSize, (8 - dstLen) / 2);
}
// Get fallback character
@ -163,7 +168,7 @@ public:
// convert at least one extra character
char bom[8];
char *dst = bom;
size_t dstSize = min((size_t)8, bomSize + *outbytesleft);
size_t dstSize = std::min((size_t)8, bomSize + *outbytesleft);
const char *src = *inbuf;
size_t srcSize = *inbytesleft;
iconv(cd, ICONV_CONST_CAST(&src), &srcSize, &dst, &dstSize);