Bump meson #57895
Bump meson #57895
Conversation
| void *ptr = realloc(old_ptr, size); | ||
| if (ptr != NULL) { | ||
| if (old_ptr != ptr) { |
There was a problem hiding this comment.
gcc was complaing about a use-after-free with old_ptr here
| @@ -59,7 +59,7 @@ def read_uint32_with_byteswap(bytes data, Py_ssize_t offset, bint byteswap): | |||
|
|
|||
| def read_uint64_with_byteswap(bytes data, Py_ssize_t offset, bint byteswap): | |||
| cdef uint64_t res | |||
| assert offset + sizeof(res) < len(data) | |||
There was a problem hiding this comment.
warning about comparison of operands with different signs (size_t is unsigned, Py_ssize_t is signed)
| @@ -137,7 +137,7 @@ except ImportError: | |||
|
|
|||
| @cython.wraparound(False) | |||
| @cython.boundscheck(False) | |||
| def memory_usage_of_objects(arr: object[:]) -> int64_t: | |||
| def memory_usage_of_objects(arr: ndarray[object]) -> int64_t: | |||
There was a problem hiding this comment.
This produces:
pandas/_libs/lib.cpython-310-x86_64-linux-gnu.so.p/pandas/_libs/lib.pyx.c:90806:12: warning: '__pyx_memview_set_object' defined but not used [-Wunused-function]
90806 | static int __pyx_memview_set_object(const char *itemp, PyObject *obj) {
| ^~~~~~~~~~~~~~~~~~~~~~~~
pandas/_libs/lib.cpython-310-x86_64-linux-gnu.so.p/pandas/_libs/lib.pyx.c:90801:20: warning: '__pyx_memview_get_object' defined but not used [-Wunused-function]
90801 | static PyObject *__pyx_memview_get_object(const char *itemp) {
| ^~~~~~~~~~~~~~~~~~~~~~~~
although I am not sure why it wouldn't let me specify the ndim here.
| @@ -105,7 +111,7 @@ cdef value_count_{{dtype}}(const {{dtype}}_t[:] values, bint dropna, const uint8 | |||
| cdef: | |||
| int64_t[::1] result_counts = np.empty(table.size + na_add, dtype=np.int64) | |||
|
|
|||
| for i in range(table.size): | |||
| for i in range(<Py_ssize_t>table.size): # cast required for 32 bit builds | |||
There was a problem hiding this comment.
We alternately could change table.size to be of type Py_ssize_t but leaving that to a separate exercise
| @@ -43,7 +43,13 @@ cdef value_count_{{dtype}}(const {{dtype}}_t[:] values, bint dropna, const uint8 | |||
| # Don't use Py_ssize_t, since table.n_buckets is unsigned | |||
| khiter_t k | |||
|
|
|||
| {{c_type}} val | |||
| # without explicit initialization the compiler detects that val could | |||
There was a problem hiding this comment.
This is extremely wonky, but the compiler is right that val would be uninitialized if n==0. I think it would take a decent refactor to make Cython generate something else that the compiler would be comfortable with
|
I believe the pypy failure is a Cython issue - see cython/cython#6099 Generally I was surprised that the amount of warnings changed with this PR. But I fixed a few / ignored others (mostly for MSVC). Would be nice to get a lot of the MSVC warnings fixed up but I think that will take a dedicated initiative. |
|
This pull request is stale because it has been open for thirty days with no activity. Please update and respond to this comment if you're still interested in working on this. |
|
Closing since there hasn't been much activity here, but feel free to reopen if/when you have time to revist |
Meson 1.4.0 adds first class support for NumPy > 2.0 detection https://mesonbuild.com/Release-notes-for-1-4-0.html#new-numpy-custom-dependency
Not an urgent upgrade but figured worth starting