SDL2

@afxgroup

Quote:

I would recommend against that.

The SDK:clib2/include/dlfcn.h file is for clib2 and is incompatible with the newlib libdl.so (while the function prototypes are the same the RTLD_#? constants have different values).

I would also not recommend using shared objects with clib2 as it will likely cause confusion as clib2 shared objects will not be compatible with newlib ones and vice versa.

@salass00

"constants have different values"

I've got a warning message about this, but it worked.

I will do some test asap

@AmigaBlitter
It worked in a sense that you could compile SDL2, but it could lead to problems if you were to compile a program, that uses SDL functions to load SObjs. Fortunately not many I know does this.

@AmigaBlitter

If you want to disable dynapi (no dlfcn.h available or some issue with it), you can hack this

http://hg.libsdl.org/SDL/file/b2fd764 ... c/dynapi/SDL_dynapi.h#l56

to zero (0).

@Capehill

thank you for the tip

@all

Regarding (window) event handling...with SDL2 we can have multiple windows. There is SDL_PumpEvents() which should update events from video subsystem but it targets the whole subsystem, as far as I understand, and not a specific window.

So I guess we will loop the video device's window list and handle the pending events per window, whenever SDL_PumpEvents() gets called, or does anyone have different ideas?

@Capehill

Why not just use a shared IDCMP port for all the SDL windows?

You can do that very easily now with WA_UserPort and you don't have to worry about signals running out then from opening too many windows.

@BSzili

Hi,

could you please verify the configure script?

The other configure script (the one in the test's dir) seems that doesn't work.

Can't find the SDL2 libraries and headers path.

Perhaps we can configure manually the PKG_CONFIG_PATH?

@AmigaBlitter
There's nothing wrong with the configure script, just specify the path where the library and headers are.

@BSzili

Thank you for the reply.

Where i specify the path?
The configure script in the test's dir shouldn't be able to find it automatically?

@salass00

Sounds sensible, and actually similar to what SDL1 does I suppose, even though with one window.

Fullscreen is tricky. I tried to avoid poking SDL_video.c but finally I had to modify CREATE_FLAGS macro to include fullscreen bit during window creation.

Otherwise SDL2 will try to change window mode to fullscreen after its creation. So this hack prevents unnecessary closing and reopening the window.

Any thoughts?

If the switching to fullscreen happens after the window creation, then something like this is probably necessary.
BTW, I noticed that you re-added the memory pool semaphore from SDL1. This is a remnant from the OS3 port and no longer necessary, since MEMF_SHARED pools are protected against concurrent access:
http://wiki.amigaos.net/wiki/Exec_Memory_Pools#Concurrent_Access

@BSzili

Nice! Less code, less bugs.

If it turns out pool will only be used for input events, maybe the sizes can also be tuned, or item pool used as suggested on the same page.

@BSzili

Quote:

The AmigaOS needs to know when it safe to another process.

Process 1 write to memory 123.
Process 2 write to memory 235
Process 1 reads from memory and gets 123?

Or is some how process 2 forces to wait for process 1 to read and write?

Or does it mean that memory pool lists are safe so that adding or freeing additional memory is thread safe.

@LiveForIt
MEMF_SHARED pools are semaphore protected.
edit: Actually it's the ASOPOOL_Protected tag.

@thread

Please support main developers that are porting SDL2 to Amiga

Capehill, afxgroup, BSzili

@thread

Please support main developers that are porting SDL2 to Amiga

Capehill, afxgroup, BSzili


upload immagini gratis

@BSzili

Quote:

It's the ASOPOOL_Protected tag that is used to specify if a pool should be protected though. The MEMF_SHARED constant is just for specifying the type of memory that is allocated.

@salass00
Ooops, you are right. Somehow I mixed the two up.