Fixed by decrementing the i_refcount variable in the correct manner -
and therefore freeing the memory only when the i_refcount successfully
went to 0.
The problem was that i_refcount is an unsigned variable, and was being
decremented twice, once erroneously by mosaic_bridge, and then again in
the picture's original pf_release function. If i_refcount started at 1,
it wrapped to the maximum unsigned value rather than -1, failing the
refcount tests in the pf_release function.
Patch Authors: Phil Roffe and David Grellscheid
Signed-off-by: Antoine Cellerier <dionoea@videolan.org>
Showing
Please register or sign in to comment