How to use uFFI with String

horrido wrote

There is shockingly very little documentation on how to use uFFI. In particular, I have the following situation...

A shared C lib has the following function:

void get_machine(char *machine) {
        uname(&uname_s);
        strcpy(machine, uname_s.machine);
}

It copies the content of a string value into the buffer passed to it (char *machine).

I want to call this function from Pharo, thus:

buffer := String new: 64.
self ffiCall: #( void get_machine(String buffer) ).

The buffer never receives the string value (it just contains the 64 initiialized spaces).

Clearly, I'm misunderstanding the proper mechanism for doing this.

[There are other functions in the shared C lib that work fine, such as:

uint get_uptime() {
        sysinfo(&info);
        return info.uptime;
}

Returning integers or floats is no problem. But passing arguments by reference seems to be verboten.]

horrido wrote

Je comprends enfin...

I accidentally came across a StackOverflow answer that suggested the String object is immutable. That makes perfect sense, as I know the C function is actually trying to write to the buffer.

That gave me the idea to use a ByteArray instead. Now, everything works!

horrido wrote

There is shockingly very little documentation on how to use uFFI. In particular, I have the following situation...

A shared C lib has the following function:

void get_machine(char *machine) {
        uname(&uname_s);
        strcpy(machine, uname_s.machine);
}

It copies the content of a string value into the buffer passed to it (char *machine).

I want to call this function from Pharo, thus:

buffer := String new: 64.
self ffiCall: #( void get_machine(String buffer) ).

The buffer never receives the string value (it just contains the 64 initiialized spaces).

Clearly, I'm misunderstanding the proper mechanism for doing this.

[There are other functions in the shared C lib that work fine, such as:

uint get_uptime() {
        sysinfo(&info);
        return info.uptime;
}

Returning integers or floats is no problem. But passing arguments by reference seems to be verboten.]

Stephane Ducasse-3 wrote

Did you read the UFFI chapter?

https://ci.inria.fr/pharo-contribution/view/Books/job/PharoBookWorkInProgress/174/artifact/book-result/UnifiedFFI/UnifiedFFI.pdf

https://ci.inria.fr/pharo-contribution/view/Books/job/PharoBookWorkInProgress/174/artifact/book-result/UnifiedFFI/


I'm  turning it into a booklet but I will not write it because I'm bad in
FFI
Now if you have corrections, suggestions, please do some pull requests

https://github.com/SquareBracketAssociates/Booklet-uFFI

Stef

On Mon, May 22, 2017 at 10:42 PM, horrido <[hidden email]> wrote:

> There is shockingly very little documentation on how to use uFFI. In
> particular, I have the following situation...
>
> A shared C lib has the following function:
>
> void get_machine(char *machine) {
>         uname(&uname_s);
>         strcpy(machine, uname_s.machine);
> }
>
> It copies the content of a string value into the buffer passed to it (char
> *machine).
>
> I want to call this function from Pharo, thus:
>
> buffer := String new: 64.
> self ffiCall: #( void get_machine(String buffer) ).
>
> The buffer never receives the string value (it just contains the 64
> initiialized spaces).
>
> Clearly, I'm misunderstanding the proper mechanism for doing this.
>
> [There are other functions in the shared C lib that work fine, such as:
>
> uint get_uptime() {
>         sysinfo(&info);
>         return info.uptime;
> }
>
> Returning integers or floats is no problem. But passing arguments by
> reference seems to be verboten.]
>
>
>
> --
> View this message in context: http://forum.world.st/How-to-
> use-uFFI-with-String-tp4947890.html
> Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com.
>
>

Ben Coman wrote

On Sat, May 27, 2017 at 9:15 PM, horrido <[hidden email]> wrote:
> Yes, I did. I found it difficult to understand. It would be nice to have some
> clear examples in the documentation, for example, really simple and common
> situations such as a C function returning an integer in a
> passed-by-reference argument.
>
> Speaking of documentation, in "Pharo by Example 50," I found this statement:
>
>> (Recall that you should set halosEnabled in the Preferences
> browser.)
>
> However, nowhere else in the book is there any reference to halosEnabled or
> Preferences browser.

Nice pick up.  Now in Settings (in a fresh 60473 image), filtering on
"halo" shows only:
 * Cycle both directions
 * Encloses fullbounds
 * Exhibits bounds
so the comment seems irrelevant

and #halosEnabled & #halosEnabled: each have one implementor class-side of Morph
with one sender from MorphTest each, and one sender from Morph>>addHalo:

and "Morph halosEnabled" ==> true,  so the statement seems unnecessary.
Would you like to contribute a pull request removing it?
https://github.com/SquareBracketAssociates/UpdatedPharoByExample/blob/master/Morphic/Morphic.pier


> Did you guys forget and leave out a chunk of the book?

Its an evolutionary book originally from Squeak.  It could well be
that some parts of the book have been cleaned
better than others to match Image changes.  Thanks for reporting, and
even better if you can correct it.

cheers -ben