diff options
Diffstat (limited to 'Modules/main.c')
-rw-r--r-- | Modules/main.c | 104 |
1 files changed, 104 insertions, 0 deletions
diff --git a/Modules/main.c b/Modules/main.c index fa2002f..24c36db 100644 --- a/Modules/main.c +++ b/Modules/main.c @@ -706,6 +706,110 @@ Py_GetArgcArgv(int *argc, wchar_t ***argv) *argv = orig_argv; } + +wchar_t* +_Py_char2wchar(char* arg) +{ + wchar_t *res; +#ifdef HAVE_BROKEN_MBSTOWCS + /* Some platforms have a broken implementation of + * mbstowcs which does not count the characters that + * would result from conversion. Use an upper bound. + */ + size_t argsize = strlen(arg); +#else + size_t argsize = mbstowcs(NULL, arg, 0); +#endif + size_t count; + unsigned char *in; + wchar_t *out; +#ifdef HAVE_MBRTOWC + mbstate_t mbs; +#endif + if (argsize != (size_t)-1) { + res = (wchar_t *)PyMem_Malloc((argsize+1)*sizeof(wchar_t)); + if (!res) + goto oom; + count = mbstowcs(res, arg, argsize+1); + if (count != (size_t)-1) { + wchar_t *tmp; + /* Only use the result if it contains no + surrogate characters. */ + for (tmp = res; *tmp != 0 && + (*tmp < 0xd800 || *tmp > 0xdfff); tmp++) + ; + if (*tmp == 0) + return res; + } + PyMem_Free(res); + } + /* Conversion failed. Fall back to escaping with surrogateescape. */ +#ifdef HAVE_MBRTOWC + /* Try conversion with mbrtwoc (C99), and escape non-decodable bytes. */ + + /* Overallocate; as multi-byte characters are in the argument, the + actual output could use less memory. */ + argsize = strlen(arg) + 1; + res = (wchar_t*)PyMem_Malloc(argsize*sizeof(wchar_t)); + if (!res) goto oom; + in = (unsigned char*)arg; + out = res; + memset(&mbs, 0, sizeof mbs); + while (argsize) { + size_t converted = mbrtowc(out, (char*)in, argsize, &mbs); + if (converted == 0) + /* Reached end of string; null char stored. */ + break; + if (converted == (size_t)-2) { + /* Incomplete character. This should never happen, + since we provide everything that we have - + unless there is a bug in the C library, or I + misunderstood how mbrtowc works. */ + fprintf(stderr, "unexpected mbrtowc result -2\n"); + return NULL; + } + if (converted == (size_t)-1) { + /* Conversion error. Escape as UTF-8b, and start over + in the initial shift state. */ + *out++ = 0xdc00 + *in++; + argsize--; + memset(&mbs, 0, sizeof mbs); + continue; + } + if (*out >= 0xd800 && *out <= 0xdfff) { + /* Surrogate character. Escape the original + byte sequence with surrogateescape. */ + argsize -= converted; + while (converted--) + *out++ = 0xdc00 + *in++; + continue; + } + /* successfully converted some bytes */ + in += converted; + argsize -= converted; + out++; + } +#else + /* Cannot use C locale for escaping; manually escape as if charset + is ASCII (i.e. escape all bytes > 128. This will still roundtrip + correctly in the locale's charset, which must be an ASCII superset. */ + res = PyMem_Malloc((strlen(arg)+1)*sizeof(wchar_t)); + if (!res) goto oom; + in = (unsigned char*)arg; + out = res; + while(*in) + if(*in < 128) + *out++ = *in++; + else + *out++ = 0xdc00 + *in++; + *out = 0; +#endif + return res; +oom: + fprintf(stderr, "out of memory\n"); + return NULL; +} + #ifdef __cplusplus } #endif |