/* Search for a prefix value in an environment file (pyvenv.cfg).
If found, copy it into the provided buffer. */
int
_Py_FindEnvConfigValue(FILE *env_file, const wchar_t *key,
wchar_t *value, size_t value_size)
{
int result = 0; /* meaning not found */
char buffer[MAXPATHLEN*2+1]; /* allow extra for key, '=', etc. */
fseek(env_file, 0, SEEK_SET);
while (!feof(env_file)) {
char * p = fgets(buffer, MAXPATHLEN*2, env_file);
if (p == NULL) {
break;
}
size_t n = strlen(p);
if (p[n - 1] != '\n') {
/* line has overflowed - bail */
break;
}
if (p[0] == '#') {
/* Comment - skip */
continue;
}
wchar_t *tmpbuffer = _Py_DecodeUTF8_surrogateescape(buffer, n);
if (tmpbuffer) {
wchar_t * state;
wchar_t * tok = wcstok(tmpbuffer, L" \t\r\n", &state);
if ((tok != NULL) && !wcscmp(tok, key)) {
tok = wcstok(NULL, L" \t", &state);
if ((tok != NULL) && !wcscmp(tok, L"=")) {
tok = wcstok(NULL, L"\r\n", &state);
if (tok != NULL) {
wcsncpy(value, tok, MAXPATHLEN);
result = 1;
PyMem_RawFree(tmpbuffer);
break;
}
}
}
PyMem_RawFree(tmpbuffer);
}
}
return result;
}
/* Decode a byte string from the locale encoding with the
surrogateescape error handler (undecodable bytes are decoded as characters
in range U+DC80..U+DCFF). If a byte sequence can be decoded as a surrogate
character, escape the bytes using the surrogateescape error handler instead
of decoding them.
Use _Py_wchar2char() to encode the character string back to a byte string.
Return a pointer to a newly allocated wide character string (use
PyMem_RawFree() to free the memory) and write the number of written wide
characters excluding the null character into *size if size is not NULL, or
NULL on error (decoding or memory allocation error). If size is not NULL,
*size is set to (size_t)-1 on memory error and (size_t)-2 on decoding
error.
Conversion errors should never happen, unless there is a bug in the C
library. */
wchar_t*
_Py_char2wchar(const char* arg, size_t *size)
{
#ifdef __APPLE__
wchar_t *wstr;
wstr = _Py_DecodeUTF8_surrogateescape(arg, strlen(arg));
if (size != NULL) {
if (wstr != NULL)
*size = wcslen(wstr);
else
*size = (size_t)-1;
}
return wstr;
#else
wchar_t *res;
size_t argsize;
size_t count;
#ifdef HAVE_MBRTOWC
unsigned char *in;
wchar_t *out;
mbstate_t mbs;
#endif
#ifndef MS_WINDOWS
if (force_ascii == -1)
force_ascii = check_force_ascii();
if (force_ascii) {
/* force ASCII encoding to workaround mbstowcs() issue */
res = decode_ascii_surrogateescape(arg, size);
if (res == NULL)
goto oom;
return res;
}
#endif
#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.
*/
argsize = strlen(arg);
#else
argsize = mbstowcs(NULL, arg, 0);
#endif
if (argsize != (size_t)-1) {
if (argsize == PY_SSIZE_T_MAX)
goto oom;
argsize += 1;
if (argsize > PY_SSIZE_T_MAX/sizeof(wchar_t))
goto oom;
res = (wchar_t *)PyMem_RawMalloc(argsize*sizeof(wchar_t));
if (!res)
goto oom;
count = mbstowcs(res, arg, argsize);
if (count != (size_t)-1) {
wchar_t *tmp;
/* Only use the result if it contains no
surrogate characters. */
for (tmp = res; *tmp != 0 &&
!Py_UNICODE_IS_SURROGATE(*tmp); tmp++)
;
if (*tmp == 0) {
if (size != NULL)
*size = count;
return res;
}
}
PyMem_RawFree(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;
if (argsize > PY_SSIZE_T_MAX/sizeof(wchar_t))
goto oom;
res = (wchar_t*)PyMem_RawMalloc(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. */
PyMem_RawFree(res);
if (size != NULL)
*size = (size_t)-2;
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 (Py_UNICODE_IS_SURROGATE(*out)) {
/* 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++;
}
if (size != NULL)
*size = out - res;
#else /* HAVE_MBRTOWC */
/* 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 = decode_ascii_surrogateescape(arg, size);
if (res == NULL)
goto oom;
#endif /* HAVE_MBRTOWC */
return res;
oom:
if (size != NULL)
*size = (size_t)-1;
return NULL;
#endif /* __APPLE__ */
}
/* search_for_exec_prefix requires that argv0_path be no more than
MAXPATHLEN bytes long.
*/
static int
search_for_exec_prefix(const _PyCoreConfig *core_config,
PyCalculatePath *calculate, wchar_t *exec_prefix)
{
size_t n;
/* If PYTHONHOME is set, we believe it unconditionally */
if (core_config->home) {
wchar_t *delim = wcschr(core_config->home, DELIM);
if (delim) {
wcsncpy(exec_prefix, delim+1, MAXPATHLEN);
}
else {
wcsncpy(exec_prefix, core_config->home, MAXPATHLEN);
}
exec_prefix[MAXPATHLEN] = L'\0';
joinpath(exec_prefix, calculate->lib_python);
joinpath(exec_prefix, L"lib-dynload");
return 1;
}
/* Check to see if argv[0] is in the build directory. "pybuilddir.txt"
is written by setup.py and contains the relative path to the location
of shared library modules. */
wcsncpy(exec_prefix, calculate->argv0_path, MAXPATHLEN);
exec_prefix[MAXPATHLEN] = L'\0';
joinpath(exec_prefix, L"pybuilddir.txt");
if (isfile(exec_prefix)) {
FILE *f = _Py_wfopen(exec_prefix, L"rb");
if (f == NULL) {
errno = 0;
}
else {
char buf[MAXPATHLEN+1];
wchar_t *rel_builddir_path;
n = fread(buf, 1, MAXPATHLEN, f);
buf[n] = '\0';
fclose(f);
rel_builddir_path = _Py_DecodeUTF8_surrogateescape(buf, n);
if (rel_builddir_path) {
wcsncpy(exec_prefix, calculate->argv0_path, MAXPATHLEN);
exec_prefix[MAXPATHLEN] = L'\0';
joinpath(exec_prefix, rel_builddir_path);
PyMem_RawFree(rel_builddir_path );
return -1;
}
}
}
/* Search from argv0_path, until root is found */
copy_absolute(exec_prefix, calculate->argv0_path, MAXPATHLEN+1);
do {
n = wcslen(exec_prefix);
joinpath(exec_prefix, calculate->lib_python);
joinpath(exec_prefix, L"lib-dynload");
if (isdir(exec_prefix)) {
return 1;
}
exec_prefix[n] = L'\0';
reduce(exec_prefix);
} while (exec_prefix[0]);
/* Look at configure's EXEC_PREFIX */
wcsncpy(exec_prefix, calculate->exec_prefix, MAXPATHLEN);
exec_prefix[MAXPATHLEN] = L'\0';
joinpath(exec_prefix, calculate->lib_python);
joinpath(exec_prefix, L"lib-dynload");
if (isdir(exec_prefix)) {
return 1;
}
/* Fail */
return 0;
}
