Hi.
The detected files are all in the tarball's Source directory.
There are no problems with mixing PyObject and PyMem memory handling,
as can be seen below.
Below is an account of the use of PyMem_* (n.b. there seem to be some
non-python memory things used as well), listed following "A:" is
functions called to allocate/resize, "F:" is followed by the methods
called to free. As some are freed in multiple code paths (typically
error during initialization vs. destructor), some use different
functions in different situations. Typically, these seem to be arrays
to be passed to the underlying libraries, while the "real" python
objects are using the PyObject_* API.
This does not seem to be causing any problems at the moment, but I
would suggest keeping this bug open (without the goal-python2.5 tag
and at normal severity) and cleaning this to use only PyMem_New and
PyMem_Del as recommended by the Python API docs. Possibly the use of
PyMem_Del should be changed to SUPERLU_FREE (possibl it needs extra
checking for !=NULL then, though, I did not check).
Kind regards
T.
Here are the details:
csr_mat.c:
SSSMatObject elements ind,val,col
A: PyMem_NEW F: PyMem_DEL, PyMem_Del
correq.c
CorrEqObject element work
A: PyMem_New F: PyMem_DEL
itsolversmodule.c
local variable work in ItSolvers_*
A: PyMem_New F: PyMem_DEL
jdsymmodule.c
local variables Q, lambda in function JDSym_jdsym
A: PyMem_New F: PyMem_DEL
ll_math.c
llColIndex members link, row, root
A: PyMem_New F: PyMem_Del
LLMathObject members col, link, val (int arrays)
A: PyMem_New, PyMem_Resize
F: PyMem_DEL, PyMem_Del
preconmodule.c:
JacobiObject members dinv, temp
A: PyMem_New F: PyMem_DEL, PyMem_Del
SSORObject members temp, temp2
A: PyMem_New F: PyMem_DEL, PyMem_Del
sss_mat.c:
SSSMatObject elements ind,val,col,diag
A: PyMem_NEW F: PyMem_DEL or PyMem_Del
superlumodule.c:
SuperLUObject elements perm_r, perm_c
A: intMalloc superlu/utils.h F: SUPERLU_FREE or PyMem_Del
umfpackmodule.c
UMFPackObject elements val,row,ind,Control,Info
A: PyMem_Malloc F: PyMem_Free
--
Thomas Viehmann, http://thomas.viehmann.net/
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
