Brad Fuller wrote:

I've been fooling with this all day.. getting frustrating!

If you get frustrated enough and decide to (instead of tricking SmartSyntaxPlugin into doing the right thing) do the hard way, here is how:

primitiveGetText | text sz oop ptr | self export: true. self var: #text type: 'char*'. self var: #ptr type: 'char*'. text := self cCode: 'GetText()'. sz := self strlen: sz. "Create new String object of appropriate size" oop := interpreterProxy instantiateClass: interpreterProxy classString indexableSize: sz. "Copy data" ptr := interpreterProxy firstIndexableField: oop. self cCode:'memcpy(ptr, text, sz)'. "Return oop" interpreterProxy pop: interpreterProxy methodArgumentCount+1. interpreterProxy push: oop.

Cheers, - Andreas

On 13-Dec-05, at 6:03 PM, Brad Fuller wrote:

Brad Fuller wrote:

...

Hello all,

I'm trying to generate a 'ulong' in Slang using the "self primitive: parameters:" method.

Don't think that is likely to work. There isn't really an unsigned long type in Smalltalk - the class Unsigned is only used in some of the strange TestOSAPlugin prims along with #asOop: which tries to make a valid positive 32bit number out of it. Take a look for references to the class Unsigned to see what I mean.

If what you want is to extract the param and treat it as an unsigned long (you might also need to think about 64bitness) then you will need to pass in a SmallInteger or possibly a LargePositiveInteger - which means using Oop - and coerce it later.

...

self primitive: 'primitivePutOutputParms' parameters: #(#SmallInteger #SmallInteger #Unsigned #Float #String).

But, I don't know how. #Float generates a double and SmallInteger generates "int". What are the other ones? Is this a dictionary? When I use #Unsigned it generates:

int parm2;

interpreterProxy->success(interpreterProxy->isKindOf (interpreterProxy->stackValue(2), "Unsigned")); parm2 = interpreterProxy->stackValue(2);

which I don't think I want, do I?

Nope. It probably ought to raise an error rather than just sticking something meaningless in there.

...

Also, does anyone know how to generate void pointer?: (void *)?

coerce it - as in "self cCoerce: fred to: 'void *' "

Further to add to this thread:

I need to return a char pointer from a primitive back to Squeak, but I can't find out how to do it in Slang. Can anyone give me pointers?

I doubt you really want to return a char pointer; what use would it be? If you really want to have the pointer - it is a real not-in-object- space pointer isn't it? - then you'd need to return it as a number and then do the right thing with it when you call some other prim that can make sense of it. For example, we pass file structure pointers up to Squeak from the FilePlugin and only look at them in prims. If you actually want the String, then you need to create a String object and copy the chars from your pointer into the String instance. Take a look at MIDIPlugin>primitiveMIDIGetPortName: where the nameObj is a createdString and the C call memcpy() is used to fill it.

Using: self returnTypeC: 'char *' works great for the return of a routine, but I need the return value to be pushed on the stack for return back to Squeak.

As mentioned above, you probably don't want the actual char * from what I can see.

what I have is this which doesn't work quite right.. (primarily because I can't quite get the whole idea of this):

primitiveGetText | text |

self var: #text declareC: 'char *text'.

ok

self primitive: 'primitiveGetText'.

ok - note you are at liberty to name the prim anything here, "self primitve: 'primFred' " would work as well. You have to remember to actually *call* primFred of course.

text := self cCode: 'GetText()'.

ok - note that you can get away with replacing "self cCode:'strlen (fred)' " with "self strlen: fred" and it has the virtue that you can implement #strlen: somewhere for simulation purposes.

^text asValue: String.

bad; this is not converting text into a String, it is assuming text is already a String object and giving you the C pointer that would be valid if you treat it as a string. Poor choice of name, maybe. More often than not you'll see the code rather more directly using "interpreterProxy firstIndexableField: nameOop" instead.

tim