@broadblues

I am not an expert, I only tried ReAction once maybe 7 years ago ... The opinion of centaurz is very close to what I intented to say.

Hey ! Nice to see facts with which we can build a list for new features.

I just had a look at ReAction examples and I use MUI because it is more high-level (and portable). It seems that ReAction uses more static things when programming (menu was given as an example).

@alfkil

So nice ! It seems that my article is helpful :)

You can check your final executable with :
readelf -d testapp

The whole amount is around 65000 Euros.

If the open hardware gives the opportunity to get a board at the end, so this is good news !

feanor : I would like to put money in this bounty but I have questions :
- Do you plan to assign yourself to the project ? Else, if we put money and nobody takes the bounty, what will happen ?
- If not all steps are done ... could we stay stopped with for example 30000 Euros consumed ?
- Do you think some partners / investors you talked to would put money in this bounty ?

@Rigo

Yes, in the same project. It is just an idea but you're right, it is possible to have this kind of behaviour with 2 projects even if I think it would be less efficient.

Prepare to receive other other suggestions when I will use it more deeply

Rigo : Just a wish that I would like in CodeBench : compilation profiles.

For example, I want to use GCC for the OS4 version of a project and vbcc for the 68k version of this project. A profile would contain the compiler name, the linker name, their options, etc. It would also help to declare a profile without debug and another one with debug.

We could have several profiles and a current one, with the possibility to switch.

@DaveP

Oh no ! Not again !

Yes, the Beagle board is fun, nice, cheap but ... I don't fully understand the love for this board. It has no Ethernet port, low storage capacity, ...

What would be the usage of such a board ?

To answer your question : "Could be running on one of these?". Yes, you can run Linux on it.

@Rigo

Thanks for your reply. We know that we have to wait for the next update coming with the OS one. I just want to remind you that CodeBench is appreciated and we are impatient to use the next version :)

For improvements, Joerg, why not use the work done by Fab for the MorphOS version to bring new features in the OS4 version. Fab is open and is ok with that, I again talked to him this week-end at the Alchimie show.

It shew us the current state of OWB for MorphOS and it is great. Working together would have a big sense here and it will bring some advantages to everybody.

@alfkil

Are you sure it is a compiler bug ? What happens ? Compiler bugs are possible but they often don't appear in usual applications.
Could you explain the problem ?

Rigo : gdb could be useful, we've used it this week-end at the Alchimie show (found my bug with a human debugger ... thank you for your suggestion Fab

Several people here were using CodeBench and we talked about it. We would like to have the new version now ... unfortunately it is not available. We are impatient to use it, CodeBench has a very high potential.

jamvm is a small JVM that works with Classpath. I think it will be easier to port that the JVM by Sun.
I tried jamvm on Linux but don't know it in depth at the moment ...

JDLL was a nice meeting with few great contacts. About OO4kids, I don't think this professor will port it itself on OS4 but I had a quick and interesting talk with him.
I can't tell much more, I will contact him soon.

But I doesn't want to make false hopes.

And I have also to work on my own projects (enhancing PointRider at the moment).

It is really nice to go in non-Amiga exclusive events. We have things to show and no complex, even in this free software meeting, with our OS is not open source.

Let's have fun !

@Dirk-B

There are other PPC 405 board that are available as second hand products (I have for exemple a DHT walnut board). PPC 405 is less powerful than 440 so there no really interest compared to a Sam440.

If we start to look at PPC evaluation boards, many are available. The problem is elsewhere ...

@Rogue

Quote:

So true ! And I would add that 7 years after the first PPC motherboards are out, people still asks for support on 68k, not realizing that PPC systems bring the same feeling with more pleasure !

Thank you for working on this project, it is very important for the platform to get it. All the best !

@ZeroG

He can speak with Hyperion but as it is not sure that the board will be built, he needs to know who is interested to make investors trust the project.
But that's true : a commitment of Hyperion in the case the project is validated would bring much more answers and interest for us.

But once the board is here, anything is possible !

@Antique

I don't know anything about the meeting and announcements but I suppose it is the great news. Don't you think it is ? What would you expect ?

First, that's nice to see that the development has been continued during the last months in parallel with Sam flex. Second, it is a new PPC projects with an interesting application field.

I am entirely satisfied by this news and I am excited about learning more about that.

I reply to myself because I think I've just found a very important information to share.

My example of shared objet libSDL_ttf.so was crashing with an ISI error. I used to compile it like that :

gcc -shared -o libSDL_ttf.so SDL_ttf.o

I followed the procedure that I found but in my case, objects to link are also shared objects so it must be :

gcc -L/SOBJS/ -use-dynld -shared -o libSDL_ttf.so SDL_ttf.o -lfreetype

Tell me if I am wrong but now that works ! We have to explicit the fact that creating the shared object, it has to be linked dynamically with another one, this is why I added -use-dynld and -lfreetype.

I hope all that will help !

I've just tried to summarize all the points that were unclear and then discussed here. Please read and comment, I would like to put all information in one place later :


-use-dynld

It allows to use shared object that will be loaded / linked at runtime. But at compilation time, note that for each path it will check first if a shared object exists, else it will use the static library.
So you might think you will use a shared object because you specified -use-dynld but it could not be the case.


-Lpath

This classic option tells to the linker where it has to search the objects.
It will look first in the SDK paths.

[source]attempt to open /SDK/newlib/lib/libfreetype.so failed
attempt to open /SDK/newlib/lib/libfreetype.a failed
attempt to open /GCC/lib/gcc/ppc-amigaos/4.2.4/newlib/lib/libfreetype.so failed
attempt to open /GCC/lib/gcc/ppc-amigaos/4.2.4/newlib/lib/libfreetype.a failed
attempt to open /SDK/local/newlib/lib/libfreetype.so failed
attempt to open /SDK/local/newlib/lib/libfreetype.a failed
attempt to open /SDK/local/common/lib/libfreetype.so failed
attempt to open /SDK/local/common/lib/libfreetype.a failed
attempt to open ./libfreetype.so failed
attempt to open ./libfreetype.a failed
attempt to open /SOBJS//libfreetype.so succeeded
-lfreetype (/SOBJS//libfreetype.so)[/source]


-lW,-rpath,<runtimepath>

This is a linker option and as indicated, it contains a runtime path (see : http://en.wikipedia.org/wiki/Rpath_%28linking%29) that will be hardcoded in the final ELF executable. It is a bad idea to use a SDK related path ! This path "may either override or supplement the system default dynamic linking search path" even if I suppose that the object is search into SOBJS: first on OS4.


Check which shared objects are used

If you compiled "myprogram" and you want to see which dependencies are shared objects, do :
readelf -d myprogram

Here is an example of results, you will see used shared objects looking at lines which type is "(NEEDED)" :

[source]> readelf -d example

Dynamic section at offset 0x5d6c contains 19 entries:
Tag Type Name/Value
0x00000001 (NEEDED) Shared library: [libSDL-1.2.so]
0x00000001 (NEEDED) Shared library: [libfreetype.so]
0x00000001 (NEEDED) Shared library: [libz.so]
0x00000001 (NEEDED) Shared library: [libgcc.so]
0x00000001 (NEEDED) Shared library: [libc.so]
0x00000004 (HASH) 0x10000e8
0x00000005 (STRTAB) 0x100051c
0x00000006 (SYMTAB) 0x100023c
0x0000000a (STRSZ) 617 (bytes)
0x0000000b (SYMENT) 16 (bytes)
0x00000015 (DEBUG) 0x0
0x00000003 (PLTGOT) 0x1015e6c
0x00000002 (PLTRELSZ) 408 (bytes)
0x00000014 (PLTREL) RELA
0x00000017 (JMPREL) 0x10007a0
0x00000007 (RELA) 0x1000788
0x00000008 (RELASZ) 432 (bytes)
0x00000009 (RELAENT) 12 (bytes)
0x00000000 (NULL) 0x0[/source]


Dependencies

If a shared object it used, its own dependencies must be shared objects too. For example, I tried to test the "libSDL_ttf.so" file I compiled but as the "libfreetype.a" in the SDK subdirectories was found before the "libfreetype.so" in SOBJS: and that can't work ! In that case, prefer the static library (SDL_ttf is small) and create link of the shared object in the path "SDK:Local/newlib/lib" so it will be found at compilation time and "libfreetype.so" will be used, as you specified -use-dynld.


At runtime

We saw that shared objects are stored in SOBJS: (what means SYS:SObjs/) and the ELF loader (elf.library) will look into the path specific with the option -rpath if used. If an object is not found, a window will appear to tell which object is missing and at the end the message "file not an executable" will be printed in a console.


Version number

With the provided shared objects, the command "version" will give no results because the version was not added. Some shared objects have a version in their name (SDL-1.2, png12, ...) and I was wondering how they are managed knowing that these libraries also exist with their plain name (SDL, png, ...). Usually .so files came with a version number, so you can have multiple version libraries and usually the one without number is a link to the last version.

@afxgroup

Quote:

Right but .so will be used only if .a is not found before on the path.

About freetype includes, I think the most clean solution would be to provide the .so file, its includes (not in freetype2/) and also the static libs of the same version.