fdo#70090: Avoid race in copy vs. modification of aRetval Sequence
If, in the original code, aRetval[0] was computed before calling new MaskPrmitive2D (as apparently happens with Clang), no COW-copy of aRetval's internals was created (as the refcount was still 1), so MaskPrimitive2D's maChildren would share the Sequence internals with aRetval, so would end up having itself as its children after the assignment. Change-Id: I145ca429fdfae2ea555b8c754b407a81b989d45b
Showing
Please
register
or
sign in
to comment