This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Fix python-interactive with Python 3.6


On 2017-01-20 11:37, Pedro Alves wrote:
On 01/20/2017 03:15 PM, Simon Marchi wrote:
Since Python 3.4, the callback installed in PyOS_ReadlineFunctionPointer
should return a value allocated with PyMem_RawMalloc instead of
PyMem_Malloc.  The reason is that PyMem_Malloc must be called with the
Python Global Interpreter Lock (GIL) held, which is not the case in the
context where this function is called.  PyMem_RawMalloc was introduced
for cases like this.


+/* Starting from Python 3.4, the result of the PyOS_ReadlineFunctionPointer + callback must be allocated with PyMem_RawMalloc rather than PyMem_Malloc. */
+
+#if PY_MAJOR_VERSION == 3 && PY_MINOR_VERSION >= 4

This will break with PY_MAJOR_VERSION 4 (at some point :-) ).

Yeah, I thought about that and concluded it would be a problem for my grandchildren ;).

Write:

 #if PY_MAJOR_VERSION > 3
     || (PY_MAJOR_VERSION == 3 && PY_MINOR_VERSION >= 4)

?

Ok.

Or add some macro that makes it a bit easier to write,
like "#if HAVE_PY_AT_LEAST(3, 4)" or
"#if GDB_PY_VERSION >= 3004" (like GCC_VERSION).
Or use Python's PY_VERSION_HEX, like
"#if PY_VERSION_HEX >= 0x03040000".

I think that's better.

+#define PyOS_ReadlineFunctionPointer_Malloc PyMem_RawMalloc
+#else
+#define PyOS_ReadlineFunctionPointer_Malloc PyMem_Malloc
+#endif

It sounds like we'll find other cases that will need
to call PyMem_RawMalloc going forward?  How about handling
this similarly to how we handle fixing up other missing
Python bits, in python-internal.h.  Like:

For now it looks like this is the only place where the Python API requires using PyMem_RawMalloc. In the other cases, we are holding the GIL, thanks to gdbpy_enter, so using the regular PyMem functions is ok. It doesn't mean there won't be other cases added in the future though...

#if python < 3.4

static inline void *
gdb_PyMem_RawMalloc (size_t n)
{
  return gdb_PyMem_Malloc (n);
}
#define PyMem_RawMalloc(n) gdb_PyMem_RawMalloc (n)

#endif

... and this is cleaner anyway.  But why not just

#if python < 3.4
#define PyMem_RawMalloc PyMem_Malloc
#endif

?

+
 /* Readline function suitable for PyOS_ReadlineFunctionPointer, which
    is used for Python's interactive parser and raw_input.  In both
    cases, sys_stdin and sys_stdout are always stdin and stdout
@@ -63,7 +73,7 @@ gdbpy_readline_wrapper (FILE *sys_stdin, FILE *sys_stdout,
   /* Detect EOF (Ctrl-D).  */
   if (p == NULL)
     {
-      q = (char *) PyMem_Malloc (1);
+      q = (char *) PyOS_ReadlineFunctionPointer_Malloc (1);

and then write PyMem_RawMalloc here.

Thanks,
Pedro Alves


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]