Commit Graph

334 Commits

Author SHA1 Message Date
Serhiy Storchaka ea525a2d1a Issue #27078: Added BUILD_STRING opcode. Optimized f-strings evaluation. 2016-09-06 22:07:53 +03:00
Victor Stinner 2250c4cb81 Issue #27830: Remove unused _PyStack_AsDict()
I forgot to remove this function, I made a mistake in my revert.
2016-09-05 11:43:18 -07:00
Victor Stinner e90bdb19f2 Issue #27830: Revert, remove _PyFunction_FastCallKeywords() 2016-08-25 23:26:50 +02:00
Victor Stinner 3f1057a4b6 method_call() and slot_tp_new() now uses fast call
Issue #27841: Add _PyObject_Call_Prepend() helper function to prepend an
argument to existing arguments to call a function. This helper uses fast calls.

Modify method_call() and slot_tp_new() to use _PyObject_Call_Prepend().
2016-08-25 01:04:14 +02:00
Victor Stinner f7507dd3e8 Issue #27830: Fix _PyObject_FastCallKeywords()
Pass stack, not unrelated and uninitialized args!
2016-08-25 01:00:31 +02:00
Victor Stinner 53cb489ed0 _PyObject_FastCallDict(): avoid _Py_CheckFunctionResult()
_PyObject_FastCallDict() only requires _Py_CheckFunctionResult() for the
slow-path. Other cases already check for the result.
2016-08-25 00:39:34 +02:00
Victor Stinner 577e1f8cb4 Add _PyObject_FastCallKeywords()
Issue #27830: Similar to _PyObject_FastCallDict(), but keyword arguments are
also passed in the same C array than positional arguments, rather than being
passed as a Python dict.
2016-08-25 00:29:32 +02:00
Victor Stinner 74319ae219 Use Py_ssize_t type for number of arguments
Issue #27848: use Py_ssize_t rather than C int for the number of function
positional and keyword arguments.
2016-08-25 00:04:09 +02:00
Victor Stinner dd69564c38 PyObject_CallMethodObjArgs() now uses fast call
Issue #27809:

* PyObject_CallMethodObjArgs(), _PyObject_CallMethodIdObjArgs() and
  PyObject_CallFunctionObjArgs() now use fast call to avoid the creation of a
  temporary tuple
* Rename objargs_mktuple() to objargs_mkstack()
* objargs_mkstack() now stores objects in a C array using borrowed references,
  instead of storing arguments into a tuple

objargs_mkstack() uses a small buffer allocated on the C stack for 5 arguments
or less, or allocates a buffer in the heap memory.

Note: this change is different than the change 0e4f26083bbb, I fixed the test
to decide if the small stack can be used or not. sizeof(PyObject**) was also
replaced with sizeof(stack[0]) since the sizeof() was wrong (but gave the same
result).
2016-08-24 01:14:54 +02:00
Victor Stinner d48af09043 Backed out changeset 0e4f26083bbb (PyObject_CallMethodObjArgs) 2016-08-24 00:59:40 +02:00
Victor Stinner e62b852605 PyObject_CallMethodObjArgs() now uses fast call
Issue #27809:

* PyObject_CallMethodObjArgs(), _PyObject_CallMethodIdObjArgs() and
  PyObject_CallFunctionObjArgs() now use fast call to avoid the creation of a
  temporary tuple
* Rename objargs_mktuple() to objargs_mkstack()
* objargs_mkstack() now stores objects in a C array using borrowed references,
  instead of storing arguments into a tuple

objargs_mkstack() uses a small buffer allocated on the C stack for 5 arguments
or less, or allocates a buffer in the heap memory.
2016-08-24 00:01:56 +02:00
Victor Stinner b900939186 _PyFunction_FastCallDict() supports keyword args
Issue #27809:

* Rename _PyFunction_FastCall() to _PyFunction_FastCallDict()
* Rename _PyCFunction_FastCall() to _PyCFunction_FastCallDict()
*  _PyFunction_FastCallDict() now supports keyword arguments
2016-08-22 23:15:44 +02:00
Victor Stinner 559bb6a713 Rename _PyObject_FastCall() to _PyObject_FastCallDict()
Issue #27809:

* Rename _PyObject_FastCall() function to _PyObject_FastCallDict()
* Add _PyObject_FastCall(), _PyObject_CallNoArg() and _PyObject_CallArg1()
  macros calling _PyObject_FastCallDict()
2016-08-22 22:48:54 +02:00
Serhiy Storchaka 54cd1969ea Issue #26984: int() now always returns an instance of exact int. 2016-08-21 20:03:08 +03:00
Victor Stinner 8a31c82093 Fix PyObject_Call() parameter names
Issue #27128: arg=>args, kw=>kwargs.

Same change for PyEval_CallObjectWithKeywords().
2016-08-19 17:12:23 +02:00
Victor Stinner 0d1a799343 Avoid call_function_tail() for empty format str
Issue #27128, PyObject_CallFunction(), _PyObject_FastCall() and callmethod():
if the format string of parameters is empty, avoid the creation of an empty
tuple: call _PyObject_FastCall() without parameters.
2016-08-19 17:04:54 +02:00
Victor Stinner 71aea8e981 PEP 7: add {...} around null_error() in abstract.c
Issue #27128.
2016-08-19 16:59:55 +02:00
Victor Stinner d042f1f5eb Cleanup callmethod()
Make callmethod() less weird: don't decrement func reference counter,
the caller is now responsible to do that.

Issue #27128.
2016-08-19 16:56:49 +02:00
Victor Stinner 64faad6e45 Cleanup call_function_tail()
Make call_function_tail() less weird: don't decrement args reference counter,
the caller is now responsible to do that. The caller now also checks if args is
NULL.

Issue #27128.
2016-08-19 16:50:49 +02:00
Victor Stinner 8880708f81 call_function_tail() uses fast call
Issue #27128: Modify call_function_tail() to use _PyObject_FastCall() when args
is not a tuple to avoid the creation of a temporary tuple.

call_function_tail() is used by:

* PyObject_CallFunction()
* PyObject_CallMethod()
* _PyObject_CallMethodId()
2016-08-19 16:44:19 +02:00
Victor Stinner 9be7e7b52f Add _PyObject_FastCall()
Issue #27128: Add _PyObject_FastCall(), a new calling convention avoiding a
temporary tuple to pass positional parameters in most cases, but create a
temporary tuple if needed (ex: for the tp_call slot).

The API is prepared to support keyword parameters, but the full implementation
will come later (_PyFunction_FastCall() doesn't support keyword parameters
yet).

Add also:

* _PyStack_AsTuple() helper function: convert a "stack" of parameters to
  a tuple.
* _PyCFunction_FastCall(): fast call implementation for C functions
* _PyFunction_FastCall(): fast call implementation for Python functions
2016-08-19 16:11:43 +02:00
Martin Panter 32d2ce3561 Issue #27581: Merge overflow fix from 3.5 2016-07-25 03:31:29 +00:00
Martin Panter e8db861f47 Issue #27581: Don’t rely on overflow wrapping in PySequence_Tuple()
Patch by Xiang Zhang.
2016-07-25 02:30:05 +00:00
Martin Panter 590dcab95a Merge spelling fixes from 3.5 2016-06-20 07:55:14 +00:00
Martin Panter 2275e626b1 Fix spelling errors in code comments 2016-06-20 07:52:50 +00:00
Serhiy Storchaka 5697c4b641 Comment fixes extracted from patch by Demur Rumed. 2016-06-12 17:02:10 +03:00
Serhiy Storchaka 16931c3559 Issue #26983: float() now always return an instance of exact float.
The deprecation warning is emitted if __float__ returns an instance of
a strict subclass of float.  In a future versions of Python this can
be an error.
2016-06-03 21:42:55 +03:00
Martin Panter 9513ba3b8f Issue #24802: Merge null termination fixes from 3.5 2015-11-07 03:15:32 +00:00
Martin Panter 61d6e4ae9d Issue #24802: Merge null termination fixes from 3.4 into 3.5 2015-11-07 02:56:11 +00:00
Martin Panter eeb896c411 Issue #24802: Copy bytes-like objects to null-terminated buffers if necessary
This avoids possible buffer overreads when int(), float(), compile(), exec()
and eval() are passed bytes-like objects. Similar code is removed from the
complex() constructor, where it was not reachable.

Patch by John Leitch, Serhiy Storchaka and Martin Panter.
2015-11-07 02:32:21 +00:00
Victor Stinner e20310fa19 Issue #25556: Add assertions to PyObject_GetItem() to ensure that an exception
is raised when it returns NULL.

Simplify also ceval.c: rely on the fact that PyObject_GetItem() raised an
exception when it returns NULL.
2015-11-05 13:56:58 +01:00
Martin Panter 2eb819f7a8 Issue #25523: Merge "a" to "an" fixes from 3.4 into 3.5 2015-11-02 04:04:57 +00:00
Martin Panter 7462b64911 Issue #25523: Correct "a" article to "an" article
This changes the main documentation, doc strings, source code comments, and a
couple error messages in the test suite. In some cases the word was removed
or edited some other way to fix the grammar.
2015-11-02 03:37:02 +00:00
Martin Panter ec1aa5c2a1 More typos in 3.5 documentation and comments 2015-10-07 11:03:53 +00:00
Raymond Hettinger 740d6134f1 merge 2015-05-17 14:47:00 -07:00
Raymond Hettinger 610a51f364 Issue #23757: Only call the concrete list API for exact lists. 2015-05-17 14:45:58 -07:00
Victor Stinner 944fbcc478 Issue #23571: Enhance _Py_CheckFunctionResult()
Too bad, sometimes Py_FatalError() is unable to write the exception into
sys.stderr (on "AMD64 OpenIndiana 3.x" buildbot, the buildbot was probably out
of memory).

Call Py_FatalError() with a different message for the two cases (result+error,
or no result and no error).
2015-03-24 16:28:52 +01:00
Victor Stinner 3b06dfb9d1 Issue #23571: In debug mode, _Py_CheckFunctionResult() now calls
Py_FatalError() instead of using an assertion in debug mode. Py_FatalError()
displays the current exception and the traceback which contain more information
than just the assertion error.
2015-03-24 12:53:59 +01:00
Victor Stinner efde146b0c Issue #23571: _Py_CheckFunctionResult() now gives the name of the function
which returned an invalid result (result+error or no result without error) in
the exception message.

Add also unit test to check that the exception contains the name of the
function.

Special case: the final _PyEval_EvalFrameEx() check doesn't mention the
function since it didn't execute a single function but a whole frame.
2015-03-21 15:04:43 +01:00
Victor Stinner 283f3f8ab3 Issue #23571: Oops, fix #ifdef assert()
assert() are noop when NDEBUG is defined. We want the opposite.
2015-03-09 15:55:37 +01:00
Victor Stinner 4a7cc88472 Issue #23571: PyObject_Call(), PyCFunction_Call() and call_function() now
raise a SystemError if a function returns a result and raises an exception.
The SystemError is chained to the previous exception.

Refactor also PyObject_Call() and PyCFunction_Call() to make them more readable.

Remove some checks which became useless (duplicate checks).

Change reviewed by Serhiy Storchaka.
2015-03-06 23:35:27 +01:00
Stefan Krah 5178d91be0 Issue #14203: Remove obsolete support for view==NULL in PyBuffer_FillInfo()
and bytearray_getbuffer().  Both functions now raise BufferError in that
case.
2015-02-03 16:57:21 +01:00
Serhiy Storchaka 3dd3e26680 Issue #22896: Avoid to use PyObject_AsCharBuffer(), PyObject_AsReadBuffer()
and PyObject_AsWriteBuffer().
2015-02-03 01:25:42 +02:00
Serhiy Storchaka 4fdb68491e Issue #22896: Avoid to use PyObject_AsCharBuffer(), PyObject_AsReadBuffer()
and PyObject_AsWriteBuffer().
2015-02-03 01:21:08 +02:00
Stefan Krah 2e6658b4a7 Merge from 3.4. 2015-02-01 19:19:49 +01:00
Stefan Krah 7213fcc27e Issue #23370: Fix off-by-one error for non-contiguous buffers. 2015-02-01 16:19:23 +01:00
Stefan Krah 363af44a4a Issue #22445: PyBuffer_IsContiguous() now implements precise contiguity
tests, compatible with NumPy's NPY_RELAXED_STRIDES_CHECKING compilation
flag.  Previously the function reported false negatives for corner cases.
2015-02-01 14:53:54 +01:00
Ethan Furman b95b56150f Issue20284: Implement PEP461 2015-01-23 20:05:18 -08:00
R David Murray 861470c836 #16518: Bring error messages in harmony with docs ("bytes-like object")
Some time ago we changed the docs to consistently use the term 'bytes-like
object' in all the contexts where bytes, bytearray, memoryview, etc are used.
This patch (by Ezio Melotti) completes that work by changing the error
messages that previously reported that certain types did "not support the
buffer interface" to instead say that a bytes-like object is required.  (The
glossary entry for bytes-like object references the discussion of the buffer
protocol in the docs.)
2014-10-05 11:47:01 -04:00
Georg Brandl 72b8a80e59 Closes #22540: speed up PyObject_IsInstance and PyObject_IsSubclass in the common case that the second argument has metaclass "type". 2014-10-03 09:26:37 +02:00