Hi guys,

The recent threat on Raspberry libraries made me think. I remember that it was possible in Delphi to access .ddl libraries written in C or C++ (a bit more work) by wrapping them and using a "far" call? I did this a while ago, so not clear on the details anymore. Typically I would write a driver etc. for hardware in C/C++ and then have a Delphi wrapper class that I could use to access the hardware. I assume this is still possible in Lazarus? Are there any tutorials etc. or will the old Delphi methods still work? Will this also work with Linux .so files? For that matter, are there any other way for Lazarus to get direct access to the hardware (except maybe use inline assembler)?

Thanks for bringing me back up to speed!

Cheers,
Chavoux
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20150215/96d108f2/attachment.html>

C libraries are no problem. In fact we have many import units for various C libraries (e.g. X11, Gtk, etc.). C++ is more problematic. In theory the compiler contains support for C++ classes, but that is nowhere near usable currently so a C++ API needs to be flattened to a C one like is done with the Qt libraries.

Regards,
Sven

lol, speaking of which Sven, what's the progress report on the C++ Classes? :D

- Dennis

"Den" wrote

lol, speaking of which Sven, what's the progress report on the C++

Classes? :D

Currently not working on it. Generics and packages have higher priority. But I plan to continue it especially since the API of our company's OS is C++ based ^^

Regards,
Sven
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20150216/cfb3449d/attachment.html>

That will be the one of the greatest features :)

zeljko

Don't look forward to it too much. It will likely only work for simple cases. E.g. once a C++ class uses operators (without providing the functionality in other ways) it will likely break down...

Regards,
Sven
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20150216/23ebc05a/attachment.html>

I thought that it'll really work. Without ability to override virtual methods it's unuseable for me.

z

Indeed.. It's no easy task for sure. Sven would have to implement all features of the C++ Class that C++ supports. I'm guessing Sven is gonna go with the basic VMT C++ support, which would probably include virtual method overriding, am I right Sven?

- Dennis

priority.

> >> But I plan to continue it especially since the API of our company's

OS

methods it's unuseable for me.

I consider virtual methods as "simple cases" ;) After all I need them as well...
What I meant in my example however are especially operators that are not supported in Pascal like "[]" or "()" or dereferenciation. Also Pascal operator overloads work differently by being static methods instead of members that can use "this"/"Self".

Regards,
Sven
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20150216/22a6b186/attachment.html>

"Den" wrote

Indeed.. It's no easy task for sure. Sven would have to implement

all features of the C++ Class that C++ supports. I'm guessing Sven is gonna go with the basic VMT C++ support, which would probably include virtual method overriding, am I right Sven?

More or less, yes. Plus support for automatic calling of con- and destructors depending on scope entry/exit for instances located on the stack. Maybe also special handling for cppclass generics to get the mangled names right.
With these most cases should be covered, but not every C++ class under the sun will be usable then.

Regards,
Sven
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20150216/01d2a780/attachment.html>

Hi again...

<snip>

From: Den

I was maybe thinking more low-level (e.g. when using a compiled C++ library without the source code)? All C++ class methods are translated into (mangled) functions in any case, so as long as Lazarus can wrap these as methods in a class of its own it should work? It should even be possible to wrap 2 or 3 C++ classes into one Lazarus class?

Cheers and thanks for the insights so far.
Chavoux
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20150216/d563080b/attachment-0001.html>

"Chavoux Luyt" wrote

I was maybe thinking more low-level (e.g. when using a compiled C++

library without the source code)? All C++ class methods are translated into (mangled) functions in any case, so as long as Lazarus can wrap these as methods in a class of its own it should work? It should even be possible to wrap 2 or 3 C++ classes into one Lazarus class?

Let me say it this way: it is possible (with non virtual functions only), but I definitely don't recommend it.

Regards,
Sven
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20150216/c64e6e90/attachment.html>