On Sat, Sep 26, 2015 at 09:08:35AM -0400, John Ferlan wrote:
>
>
> On 09/24/2015 10:01 AM, Pavel Hrdina wrote:
> > Signed-off-by: Pavel Hrdina <phrdina@xxxxxxxxxx>
> > ---
> > libvirt-override.c | 9 +++------
> > 1 file changed, 3 insertions(+), 6 deletions(-)
> >
> > diff --git a/libvirt-override.c b/libvirt-override.c
> > index 14aa0e9..114104b 100644
> > --- a/libvirt-override.c
> > +++ b/libvirt-override.c
> > @@ -2303,12 +2303,9 @@ libvirt_virRegisterErrorHandler(ATTRIBUTE_UNUSED PyObject * self,
> > pyobj_f);
> >
> > virSetErrorFunc(NULL, libvirt_virErrorFuncHandler);
> > - if (libvirt_virPythonErrorFuncHandler != NULL) {
> > - Py_XDECREF(libvirt_virPythonErrorFuncHandler);
> > - }
> > - if (libvirt_virPythonErrorFuncCtxt != NULL) {
> > - Py_XDECREF(libvirt_virPythonErrorFuncCtxt);
> > - }
> > +
> > + Py_XDECREF(libvirt_virPythonErrorFuncHandler);
> > + Py_XDECREF(libvirt_virPythonErrorFuncCtxt);
>
> I keep looking at this and thinking why? Why was it added and what is
> it protecting. Looking at libvirt_virErrorFuncHandler it seems only
> libvirt_virPythonErrorFuncCtxt gets the Py_XINCREF and that's based on
> whether libvirt_virPythonErrorFuncHandler is NULL or not.
I don't understand what do you mean by this??
>
> Just feels like something subtle is going on. Also it's a register
> handler and we're decrementing something that never got incremented.
Check the whole function. It will decrement the old handlers, check for
existence of new handlers and clean/set new handlers. And before setting the
new handlers, we need to increment them, otherwise they will be cleared by GC.
Everything that is parsed by 'PyArg_ParseTuple' has to be incremented, if you
need to use those object outside of the function scope, where you've called that
parser.
Pavel
>
> John
> >
> > if ((pyobj_f == Py_None) && (pyobj_ctx == Py_None)) {
> > libvirt_virPythonErrorFuncHandler = NULL;
> >
--
libvir-list mailing list
libvir-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/libvir-list