exceptions, Python, Cython, C

[Robert Miller]

I'm working on wrapping a C package in Cython for use in Sage, and I'm
having some trouble. I think the problem has to do with the fact that
the C program uses stderr to report issues. When the program exits, I
get the following message:

SystemError: error return without exception set

I've looked around online, and found a lot of posts about this issue
for specific C packages, where someone is trying to use the C API to
wrap some C package in Python. But none of the solutions I found
seemed relevant. I also looked in the Cython documentation, but it
doesn't have quite enough information to help. (Please prove me wrong
by posting a link!)

When the code is run, it doesn't seem to go down a path which actually
uses stderr, but it does successfully use stdout to report its result,
before raising the error. It does make it all the way to the "return
0;" line, but doesn't seem to actually return to the Cython function
I'm calling it from. Also, I can't use try/except to catch the
SystemError.

[Robert Bradshaw]

This usually means you're returning NULL for an object, or perhaps an
error code for a special method (without setting an error). Perhaps
some code would be helpful?

- Robert

[Nils Bruin]

On Feb 24, 8:52 pm, Robert Miller <rlmills...@gmail.com> wrote:
> I'm working on wrapping a C package in Cython for use in Sage, and I'm
> having some trouble. I think the problem has to do with the fact that
> the C program uses stderr to report issues. When the program exits, I
> get the following message:
>
> SystemError: error return without exception set

I doubt that sending something or not sending something to stderr is
going to make any difference. It's just a writable stream. My guess is
that a routine that is expected to return a PyObject* is returning
NULL. This signals the occurrence of an error in Python and I can
imagine that when the appropriate exception reporting fields are not
set up when this happens, this is what one would expect.

I imagine this could happen if cython thinks a routine returns a
PyObject * (i.e., object -- the default type in cython), but the
routine in reality returns an integer 0. A mistranslated ".h" file
could easily cause this.

[Tom Boothby]

FWIW, I get that error message on the following:

sage: cython("""
cdef int foo(int z) except -1:
return z

def bar(z):
return foo(z)
""")
sage: bar(-1)
---------------------------------------------------------------------------
SystemError Traceback (most recent call last)

/home/tom/sage-4.6.2.rc0/<ipython console> in <module>()

[Robert Bradshaw]

Yep. You have to write "except? -1" if -1 is a valid non-error return
value. http://docs.cython.org/src/userguide/language_basics.html#error-return-values

- Robert

[Jason Grout]

Note that Robert M. said it made it down to a "return 0" line. So if
Cython is expecting a pointer back, but the program does "return 0",
then that would be equivalent to returning the NULL pointer, right?

Hopefully I'm not confusing the issue...

Thanks,

Jason

[Robert Bradshaw]

Yes, but NULL is only the error indicator for an object return type,
in which case 0 would be converted to a Python int.

Is the function in question declared as "except 0"?

- Robert

[Robert Miller]

The problem: I had a C function declared as "int foo()" but was
importing it in Cython as "cdef foo()". Thus when foo returned 0,
Cython thought this was a NULL pointer and raised the error above.

Thanks for the help, everyone.

[Robert Bradshaw]

That would do it :). This is why we need automatic .h -> .pxd
declarations. Also, it's been thrown around a couple of times that it
may be worth making "cdef foo()" a warning, forcing the user to
explicitly declare "cdef object foo()" (or, in your case, "cdef int
foo() :-).

> --
> To post to this group, send an email to sage-...@googlegroups.com
> To unsubscribe from this group, send an email to sage-devel+...@googlegroups.com
> For more options, visit this group at http://groups.google.com/group/sage-devel
> URL: http://www.sagemath.org
>