Mercurial > cpython-lru_cache-weakref
view Python/mysnprintf.c @ 33940:dfd07b7be407 legacy-trunk
SF bug 1061968: threads: segfault or Py_FatalError at exit
PyGILState_Ensure(): The fix in 2.4a3 for bug 1010677 reintroduced thread
shutdown race bug 225673. Repaired by (once again) ensuring the GIL is
held whenever deleting a thread state.
Alas, there's no useful test case for this shy bug. Four years ago, only
Guido could provoke it, on his box, and today only Armin can provoke it
on his box. I've never been able to provoke it (but not for lack of
trying!).
This is a critical fix for 2.3.5 too, since the fix for 1010677 got
backported there already and so also reintroduced 225673. I don't intend to
backport this fix. For whoever (if anyone) does, there are other thread
fixes in 2.4 that need backporting too, and I bet they need to happen first
for this patch to apply cleanly.
| author | Tim Peters <tim.peters@gmail.com> |
|---|---|
| date | Mon, 08 Nov 2004 04:30:21 +0000 |
| parents | 8af86a952b2c |
| children | 2925b568aacc a8cc8e6f0c1a |
line wrap: on
line source
#include "Python.h" #include <ctype.h> /* snprintf() wrappers. If the platform has vsnprintf, we use it, else we emulate it in a half-hearted way. Even if the platform has it, we wrap it because platforms differ in what vsnprintf does in case the buffer is too small: C99 behavior is to return the number of characters that would have been written had the buffer not been too small, and to set the last byte of the buffer to \0. At least MS _vsnprintf returns a negative value instead, and fills the entire buffer with non-\0 data. The wrappers ensure that str[size-1] is always \0 upon return. PyOS_snprintf and PyOS_vsnprintf never write more than size bytes (including the trailing '\0') into str. If the platform doesn't have vsnprintf, and the buffer size needed to avoid truncation exceeds size by more than 512, Python aborts with a Py_FatalError. Return value (rv): When 0 <= rv < size, the output conversion was unexceptional, and rv characters were written to str (excluding a trailing \0 byte at str[rv]). When rv >= size, output conversion was truncated, and a buffer of size rv+1 would have been needed to avoid truncation. str[size-1] is \0 in this case. When rv < 0, "something bad happened". str[size-1] is \0 in this case too, but the rest of str is unreliable. It could be that an error in format codes was detected by libc, or on platforms with a non-C99 vsnprintf simply that the buffer wasn't big enough to avoid truncation, or on platforms without any vsnprintf that PyMem_Malloc couldn't obtain space for a temp buffer. CAUTION: Unlike C99, str != NULL and size > 0 are required. */ int PyOS_snprintf(char *str, size_t size, const char *format, ...) { int rc; va_list va; va_start(va, format); rc = PyOS_vsnprintf(str, size, format, va); va_end(va); return rc; } int PyOS_vsnprintf(char *str, size_t size, const char *format, va_list va) { int len; /* # bytes written, excluding \0 */ #ifndef HAVE_SNPRINTF char *buffer; #endif assert(str != NULL); assert(size > 0); assert(format != NULL); #ifdef HAVE_SNPRINTF len = vsnprintf(str, size, format, va); #else /* Emulate it. */ buffer = PyMem_MALLOC(size + 512); if (buffer == NULL) { len = -666; goto Done; } len = vsprintf(buffer, format, va); if (len < 0) /* ignore the error */; else if ((size_t)len >= size + 512) Py_FatalError("Buffer overflow in PyOS_snprintf/PyOS_vsnprintf"); else { const size_t to_copy = (size_t)len < size ? (size_t)len : size - 1; assert(to_copy < size); memcpy(str, buffer, to_copy); str[to_copy] = '\0'; } PyMem_FREE(buffer); Done: #endif str[size-1] = '\0'; return len; }
