I don't know what is going on in your case, but just so folks are clear on the difference between po
& p
:
The p
command (a.k.a. expr --
) takes the arguments it is given, compiles them as though they were a source code expression written in the context of the current frame, executes the result - either by running an interpreter on the result of the compilation if that is possible, or by JITing the result of the compilation, inserting it into the target program, and running it there. Then it prints the result of the evaluation.
The po
command (a.k.a. expr --O --
) does everything that p
does, but instead of printing the result, if the result is a pointer to an ObjC object, it calls that object's "description" method, and prints the string returned by that method(*). Similarly, if the result is a CF object, it will call CFShow and print the result of that. If both these attempts fail, it will go ahead and print the result as p
would have.
So po
is mostly like p
. But you can get some weird results if you use po
on things that it aren't actually objects. For instance, ObjC has a optimization (tagged pointers) that represent the contents of some objects (e.g. NSNumbers) in the object pointer. There isn't a "real" object, just a cooked pointer. So if you try to po
an integer that just happens to look like a tagged pointer, you'll get the description of some probably unrelated ObjC object, not the value of the integer.
And of course, po
is doing a lot more work, so unless you really want some object's description of itself, p
is more efficient.
- Actually, it calls debugDescription, if that exists, and falls back to description if it doesn't...
po
equivalent toexpr -O --
(single dash vs double dash)? – Ret