On 18 Sep 2015, at 05:55, Ryan Macnak <rmacnak@gmail.com> wrote:

On Tue, Sep 15, 2015 at 8:29 AM, Esteban Lorenzano <estebanlm@gmail.com> wrote:

Hi,

I’m working on merge FFI with Alien as a way to provide a single callback mechanism (I already talked with Eliot and he agrees this is the way to go in the future).
That means review certain stuff, a couple of them I already submitted in my last commit.
Now I have this problem. This is my example method:

exampleCqsort
        | cb rand values orig sort |

        rand := Random new.
        values := FFIExternalArray newType: 'double' size: 100.
        1 to: 100 do: [ :i| values at: i put: rand next ].
        orig := (1 to: 100) collect: [:i| values at: i].
        cb := Callback
                signature:  #(int (*)(const void *, const void *))
                block: [ :arg1 :arg2 | ((arg1 doubleAt: 1) - (arg2 doubleAt: 1)) sign ].
        [
                self primQsort: values getHandle with: 100 with: values type typeSize with: cb thunk.
                sort := values asArray ]
        ensure: [ values free ].
        ^orig -> sort

primQsort: array with: count with: size with: compare
        <cdecl: void 'qsort' (void* ulong ulong void*) module: 'libc’>
        ^ self primitiveFailed

… that basically works. But this depends on this method in Callback:

voidstarvoidstarRetint: callbackContext sp: spAlien
        <signature: #(int (*)(const void *, const void *)) abi: 'IA32'>
        | value |

        value := block
                        value: (Alien forPointer: (spAlien unsignedLongAt: 1))
                        value: (Alien forPointer: (spAlien unsignedLongAt: 5)).
        ^ callbackContext wordResult: value

And I want to remove Alien from the equation (there will still be some… and there will be the primitives of course, but I want to reduce their use at the minimum for obvious reasons: hide the extra complexity to users).
So this function would be something like this instead:

voidstarvoidstarRetint: callbackContext sp: spAlien
        <signature: #(int (*)(const void *, const void *)) abi: 'IA32'>
        | value |

        value := block
                        value: (spAlien pointerAt: 1) “It will answer an ExternalAddress”
                        value: ((spAlien pointerAt: 5). “It will answer an ExternalAddress”
        ^ callbackContext wordResult: value

 So… this *should* work but it doesn’t because in primitiveFFIDoubleAt: I have a call to

ffiAddressOf: rcvr startingAt: byteOffset size: byteSize

who mades this check:

                "don't you dare to read from object memory!"
                (addr == 0 or:[interpreterProxy isInMemory: addr])
                        ifTrue:[^interpreterProxy primitiveFail].

and well… since I’m calling things from a callback the #isInMemory: method fails.

If I took out that method everything works ok and since the equivalent IA32ABIPlugin does not have that check I can suppose it is ok to remove the validation.

But! Since there was that comment ("don't you dare to read from object memory!”)… I’m afraid I can be making a mistake.

What do you think?

The elements of your FFIExternalArray should be in malloc's heap, not in the Smalltalk heap, so the check should pass.