Sign in to follow this  
UltimateWalrus

Marmalade: Angelscript runs fine on x86, crashes on ARM

Recommended Posts

WitchLord    4677
Great to hear it is finally working.

Do let me know if you encounter other problems.

I'll probably have to include a Marmalade makefile for the AngelScript library, especially since it is not so straightforward to figure out how to configure it.

Share this post


Link to post
Share on other sites
UltimateWalrus    229
Yep, it's working, and I'm finally able to run AngelScript on device using CDECL calling conventions. :]

Thanks a ton for all your help. Attached is a complete Marmalade project that implements "Hello World" by calling a print function from AngelScript, and runs just fine on devices without the asCALL_GENERIC workaround. Should save some time if anyone else needs to get AngelScript/Marmalade working in the future. You're welcome to reference it if you wanna put Marmalade makefiles in the AngelScript library.

Share this post


Link to post
Share on other sites
Braden M    126
Hi everyone,

I want to say a big thank you to all of you for your hard work in getting AngelScript to work on Marmalade. Unaware of UltimateWalrus's latest posts, I spent some time last week integrating AngelScript into Marmalade, and have put together yet another integration of AngelScript and Marmalade: [url="https://github.com/mstarinc/angelscript-marmalade"][b]angelscript-marmalade[/b] project on GitHub[/url].

It seems to be working well on MSVC/x86, VS/GCC-ARM, XCode/Clang/x86, XCode/GCC/x86, XCode/ARM, and Scons/ARM. I've included portable binaries in the distribution so that it can be used right away. If one does want to [re-]compile it, it should compile easily and run with native calling conventions on Marmalade Windows, Marmalade Mac, or Marmalade ARM. (Although due to the presence of the ".S" files, one line of the MKB must be commented out before it will compile with Visual Studio). I made sure that the AngelScript code is completely unmodified; in fact, if you use the MKB to build the project, it will download the unmodified source upstream code as part of the build process.

I'm just getting started with AngelScript, but everything looks very promising so far. Just wanted share my code and to say thanks [img]http://public.gamedev.net//public/style_emoticons/default/smile.png[/img]

Share this post


Link to post
Share on other sites
WitchLord    4677
Hi Braden,

thanks for sharing your work and for letting us know that it works on all these different platforms.

There are now quite a few projects using AngelScript with Marmalade. [img]http://public.gamedev.net//public/style_emoticons/default/smile.png[/img]

I see that you use a couple of predefines to determine the target platform, i.e. I3D_ARCH_X86 and I3D_ARCH_ARM. How did you find out about these? It looks like something I can use directly within the as_config.h so it wouldn't be necessary to define anything other than MARMALADE from the project files.

Actually perhaps I can even use the existance of I3D_ARCH_X86 or I3D_ARCH_ARM to identify that the library is being compiled for Marmalade, in which case it wouldn't be necessary to define anything at all from the Marmalde makefile.

Regards,
Andreas Edited by Andreas Jonsson

Share this post


Link to post
Share on other sites
Braden M    126
That is an excellent idea: You could use (I3D_ARCH_X86 || I3D_ARCH_ARM) as a test for the Marmalade platform. It would be nice to remove the predefines from the MKB makefile if possible. I found out about these defines from the Marmalade forum; they are not in the documentation, although they have been around for a long time. (Unfortunately there are quite a few of these little Marmalade "tricks" that are completely undocumented except for the occasional insightful forum post from Marmalade staff.)

Edit: It's worth mentioning that Marmalade also supports other platforms (although they're not nearly as popular as x86 and ARM); their preprocessor definitions are: I3D_ARCH_AMD64, I3D_ARCH_MIPS, I3D_ARCH_PS3, I3D_ARCH_X360, and I3D_ARCH_PSP. I've never worked with any of those other platforms. The full list of symbols can be seen in the "temp_defines_debug.h" file that Marmalade's makefile_builder creates during the build process.

Edit 2: Upon looking at that file further, I see that you could also detect the Marmalade platform by looking for "IW_SDK", "__S3E__", or any number of other definitions. Check out [url="https://gist.github.com/4211333"]this copy of an example temp_defines_debug.h[/url] file to see. Edited by Braden M

Share this post


Link to post
Share on other sites
WitchLord    4677
Great. It looks like __S3E__ is a useful define to look for to identify Marmalade. With this it should no longer be necessary to manually define Marmalade anywhere. I3D_ARCH_ARM and I3D_ARCH_X86 etc will be used to determine for which CPU the library is compiled, and will control the code for the native calling conventions.

I'll make the necessary changes in the as_config.h, and let you know when it's checked in.

Share this post


Link to post
Share on other sites
WitchLord    4677
I've checked in the changes in revision 1490.

It should no longer be necessary to make any defines to compile AngelScript for Marmalade. Please give it a try and let me know if there are any problems.

Based on the tests that UltimateWalrus had made, I configured Marmalade to use the same ABI as is used for Android. It's possible that later on you'll find some functions that don't work properly if this assumption is incorrect. If that happens, please tell me and I'll make the necessary changes in the library.

Share this post


Link to post
Share on other sites
Braden M    126
Thanks Andreas! I really appreciate how responsive you are with fixes to AngelScript. I just tried out r1490 and it works perfectly as far as I can tell. I've put the new MKB in [url="https://github.com/mstarinc/angelscript-marmalade/tree/angelscript2.26"]a separate branch[/url] for now, and I'll update "master" once 2.26 is released. The only "tweak" that I had to leave in was a definition of "fopen_s" that I put in preinclude.h, because Marmalade's standard library doesn't have that function available, even in Visual Studio.

Also, FYI, there are two compiler warnings that come up with Marmalade's ARM GCC4.4 (arm-none-eabi-g++); the first one looks like it could possibly affect bytecode output so I'll mention it in case you weren't aware of it.

[CODE]
angelscript/source/as_bytecode.cpp: In member function 'void asCByteCode::DebugOutput(const char*, asCScriptEngine*, asCScriptFunction*)':
angelscript/source/as_bytecode.cpp:2114: warning: format '%3d' expects type 'int', but argument 3 has type 'long unsigned int'


add_on/scripthelper/scripthelper.cpp: In function 'int WriteConfigToFile(asIScriptEngine*, const char*)':
add_on/scripthelper/scripthelper.cpp:194: warning: format '%X' expects type 'unsigned int', but argument 3 has type 'asDWORD'
[/CODE]
The second warning occurs a total of ten times at different lines in scripthelper.cpp.

Share this post


Link to post
Share on other sites
WitchLord    4677
Thanks, these last issues should now be fixed in revision 1493.

You should not need the fopen_s tweak anymore. I had forgotten a few places were I was using fopen_s. These are in either case only used when compiling the library in debug mode.

Share this post


Link to post
Share on other sites
Braden M    126
Hi Andreas, I just found two more places where fopen_s is still causing that error even with the latest revision (if I omit the preinclude.h file) - both in add_ons:

scriptfile/scriptfile.cpp (line 320)
scripthelper/scripthelper.cpp (line 169)

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this