🔒 Locked 🎮 Unity

TestReturn failed

Started by Szpak Dec 11, 2004 at 1:39 PM 7 replies 2.1k views

Original Post

Dec 11, 2004 at 1:39 PM

Hi, I downloaded and compiled AngelScript 1.10.1b with Dev-Cpp 4.9.9.1 (MinGW 3.3.1). test_build_performance worked fine, but test_feature returned: TestReturn (float): cfunction didn't return properly. Expected 18.870001, got 0.000000 I don't know is it related, but I had few warnings. Is this version compatible with MinGW? Thanks for help. g++.exe -c ../../source/test_constructor.cpp -o obj/test_constructor.o -I"E:/Dev-Cpp/include/c++/3.3.1" -I"E:/Dev-Cpp/include/c++/3.3.1/mingw32" -I"E:/Dev-Cpp/include/c++/3.3.1/backward" -I"E:/Dev-Cpp/lib/gcc-lib/mingw32/3.3.1/include" -I"E:/Dev-Cpp/include" -I"E:/Dev-Cpp/Common/Include" -I"../../../angelscript/include" -fsave-memoized -fexpensive-optimizations -O2 ../../source/test_constructor.cpp: In function bool TestConstructor()': ../../source/test_constructor.cpp:66: warning: invalid access to non-static data member CTestConstructor::a' of NULL object ../../source/test_constructor.cpp:66: warning: (perhaps the

offsetof' macro was used incorrectly) ../../source/test_constructor.cpp:67: warning: invalid access to non-static data member

CTestConstructor::b' of NULL object ../../source/test_constructor.cpp:67: warning: (perhaps the

offsetof' macro was used incorrectly) g++.exe -c ../../source/test_vector3.cpp -o obj/test_vector3.o -I"E:/Dev-Cpp/include/c++/3.3.1" -I"E:/Dev-Cpp/include/c++/3.3.1/mingw32" -I"E:/Dev-Cpp/include/c++/3.3.1/backward" -I"E:/Dev-Cpp/lib/gcc-lib/mingw32/3.3.1/include" -I"E:/Dev-Cpp/include" -I"E:/Dev-Cpp/Common/Include" -I"../../../angelscript/include" -fsave-memoized -fexpensive-optimizations -O2 ../../source/test_vector3.cpp: In function

bool TestVector3()': ../../source/test_vector3.cpp:34: warning: invalid access to non-static data member Vector3::x' of NULL object ../../source/test_vector3.cpp:34: warning: (perhaps the offsetof' macro was used incorrectly) ../../source/test_vector3.cpp:35: warning: invalid access to non-static data member Vector3::y' of NULL object ../../source/test_vector3.cpp:35: warning: (perhaps the offsetof' macro was used incorrectly) ../../source/test_vector3.cpp:36: warning: invalid access to non-static data member Vector3::z' of NULL object ../../source/test_vector3.cpp:36: warning: (perhaps the offsetof' macro was used incorrectly) g++.exe -c testvirtualinheritance.cpp -o obj/testvirtualinheritance.o -I"E:/Dev-Cpp/include/c++/3.3.1" -I"E:/Dev-Cpp/include/c++/3.3.1/mingw32" -I"E:/Dev-Cpp/include/c++/3.3.1/backward" -I"E:/Dev-Cpp/lib/gcc-lib/mingw32/3.3.1/include" -I"E:/Dev-Cpp/include" -I"E:/Dev-Cpp/Common/Include" -I"../../../angelscript/include" -fsave-memoized -fexpensive-optimizations -O2 ../../source/testvirtualinheritance.cpp: In function

bool TestVirtualInheritance()': ../../source/testvirtualinheritance.cpp:69: warning: pointer to member cast via virtual base

CVBase1' of CDerivedVirtual' ../../source/testvirtualinheritance.cpp:76: warning: pointer to member cast via virtual base CVBase2' of `CDerivedVirtual'

WitchLord

Mod

1,118

WitchLord

Dec 12, 2004 at 6:39 AM

I compile and test the library with MinGW with every release. Though I suspect that they changed something in the compiler with the latest release that make the return of floats and doubles fail in AngelScript. I've received a report of the same behaviour on Linux GNUC compiler.

I'll download the latest MinGW compiler and see if I can figure out what's going on.

The warnings you are receiving are unrelated to your problem. They are only warnings and the tests works as supposed.

WitchLord

Mod

1,118

WitchLord

Dec 13, 2004 at 8:04 AM

I just downloaded Dev-C++ 4.9.9.1 and ran the test_feature with it. Everything worked just fine, including the TestReturnF test.

Maybe you set some configurations that I didn't. Could you send me the project files for AngelScript and test_feature that you used? Or maybe tell me what you changed from the default settings.

WitchLord

Mod

1,118

WitchLord

Dec 13, 2004 at 10:00 AM

I managed to reproduce the problem. When I configured Dev-C++ to compile the library with optimizations the TestReturnF test failed.

So, until I can get this fixed I recommend that you turn off optimizations for the library.

WitchLord

Mod

1,118

WitchLord

Dec 13, 2004 at 11:17 AM

I found and fixed the bug. You'll have to replace the implementation for GetReturnedFloat() and GetReturnedDouble() in as_callfunc_x86.cpp:

asDWORD GetReturnedFloat(){    asDWORD f;#if defined ASM_INTEL	// Get the float value from ST0    __asm fstp dword ptr [f]#elif defined ASM_AT_N_T	asm("fstps %0 \n" : "=m" (f));#endif	return f;}asQWORD GetReturnedDouble(){    asQWORD d;#if defined ASM_INTEL	// Get the double value from ST0    __asm fstp qword ptr [d]#elif defined ASM_AT_N_T	asm("fstpl %0 \n" : "=m" (d));#endif	return d;}

I'd like to credit you for reporting the bug. Would you mind telling me your real name?

Szpak

Author

Szpak

Dec 15, 2004 at 8:16 AM

At first sorry for delay, but I didn't get mail notification about your replies.

Glad to help. After "patch" this test works ok. My real name is Marcin Zajaczkowski.

I have one more question. In test_multithread I had to change

#include "../../add_on/stdstr/stdstr.h"

#include "../../add_on/std_string/stdstring.h"

in utils.h to compile. I'm not sure if stdstr.h and stdstring.h is the same, but I couldn't find it in tarball and "Utils & add-ons".
After this I have

TestSharedString: Shared strings don't work as they should

I don't know is it related with this include.

Regards
Marcin

WitchLord

Mod

1,118

WitchLord

Dec 15, 2004 at 9:06 AM

Thanks for replying Marcin.

The stdstr and the stdstring are the same source. I forgot to update the multithreaded test though, thanks for letting me know.

In order for the multithreaded test to work, the AS library has to be compiled with the preprocessor flag USE_THREADS defined. Description on this and other flags can be found in as_config.h.

Regards,
Andreas

Szpak

Author

Szpak

Dec 15, 2004 at 2:56 PM

USE_THREADS helps.
Has this option any side effect? Or it's disabled by default due to portability?

Thanks for help
Marcin

WitchLord

Mod

1,118

WitchLord

Dec 16, 2004 at 10:45 AM

The main reason it is disabled by default is because the support for multithreading hasn't been fully tested.

There is a slight overhead when turning on multithreaded support because of need to use critical sections to protect the reference counting. But the actual execution of the script code shouldn't be affected.

The multithreaded support is also only available on Windows.

Topic Locked

This topic has been locked by a moderator. New replies are not allowed.