The patch adds two lines to call_find_module() to check for errors after
calling PyTokenizer_FindEncoding() since it returns NULL as a default
value so a call to PyErr_Occurred() is needed.

In PyTokenizer_FindEncoding(), PyMem_MALLOC may return NULL.
Another patch attached.

If PyMem_MALLOC() returns NULL, shouldn't MemoryError be set?

Right, here is an updated patch

Your patch looks fine, Amaury.

I think this patch should be reverted. It causes a linking error because
pgen isn't linked to libpython; it doesn't find the _PyErr_NoMemory symbol.

pgen does not exist on Windows...

What if I simply remove the call to PyErr_NoMemory?
It is not strictly necessary: the function already returns NULL without
an exception set, for example when the file cannot be opened.

On Thu, Sep 4, 2008 at 5:48 PM, Amaury Forgeot d'Arc
<report@bugs.python.org> wrote:
>
> Amaury Forgeot d'Arc <amauryfa@gmail.com> added the comment:
>
> pgen does not exist on Windows...
>
> What if I simply remove the call to PyErr_NoMemory?
> It is not strictly necessary: the function already returns NULL without
> an exception set, for example when the file cannot be opened.

That sounds ok. I can't remember what the rest of the parse does when
there's no memory, though.

>
> _______________________________________
> Python tracker <report@bugs.python.org>
> <http://bugs.python.org/issue3773>
> _______________________________________
>

You could also surround the PyErr_NoMemory with #ifndef PGEN.

committed r66224 + r66225.