Thanks. I'll try to reproduce the problem with this information.

Unfortunately it was not enough. I tried the following:

// Test repeated save/loads with shared interfaces and funcdefs
// http://www.gamedev.net/topic/656784-wrong-bytecode-with-funcdef-in-shared-interface/
{
  engine = asCreateScriptEngine(ANGELSCRIPT_VERSION);
  engine->SetMessageCallback(asMETHOD(COutStream, Callback), &out, asCALL_THISCALL);
 
  mod = engine->GetModule(0, asGM_ALWAYS_CREATE);
  mod->AddScriptSection("test",
    "funcdef void CALLBACK(); \n"
    "shared interface A \n"
    "{ \n"
    "    void Func(CALLBACK@ callback); \n"
    "} \n"
    "class B : A \n"
    "{ \n"
    "    void Func(CALLBACK@ callback){} \n"
    "} \n");
  r = mod->Build();
  if( r < 0 )
    TEST_FAILED;
 
  CBytecodeStream stream(__FILE__"shared");
  r = mod->SaveByteCode(&stream);
  if( r < 0 )
    TEST_FAILED;
 
  asDWORD crc1 = ComputeCRC32(&stream.buffer[0], asUINT(stream.buffer.size()));
 
  r = mod->LoadByteCode(&stream);
  if( r < 0 )
    TEST_FAILED;
 
  CBytecodeStream stream2(__FILE__"shared2");
  r = mod->SaveByteCode(&stream2);
  if( r < 0 )
    TEST_FAILED;
 
  asDWORD crc2 = ComputeCRC32(&stream2.buffer[0], asUINT(stream2.buffer.size()));
 
  if( crc1 != crc2 )
    TEST_FAILED;
 
  r = mod->LoadByteCode(&stream2);
  if( r < 0 )
    TEST_FAILED;
 
  engine->Release();
}

I also tried loading the bytecode into a different module, or even a different engine, but none of the tests reproduced the problem.

If you can figure out in what part of asCWriter that the output is different in the second save, it might help me understand exactly what is causing it.

Still, no luck.

I updated the test with like this:

// Test repeated save/loads with shared interfaces and funcdefs
// http://www.gamedev.net/topic/656784-wrong-bytecode-with-funcdef-in-shared-interface/
{
  engine = asCreateScriptEngine(ANGELSCRIPT_VERSION);
  engine->SetMessageCallback(asMETHOD(COutStream, Callback), &out, asCALL_THISCALL);
 
  asIScriptModule *mod1 = engine->GetModule(0, asGM_ALWAYS_CREATE);
  mod1->AddScriptSection("test",
    "funcdef void CALLBACK(); \n");
  r = mod1->Build();
  if( r < 0 )
    TEST_FAILED;
 
  asIScriptModule *mod2 = engine->GetModule(0, asGM_ALWAYS_CREATE);
  mod2->AddScriptSection("test",
    "funcdef void CALLBACK(); \n"
    "void Foo1(CALLBACK@){} \n"
    "void Foo2(){Foo1(null);} \n");
  r = mod2->Build();
  if( r < 0 )
    TEST_FAILED;
 
  CBytecodeStream stream1(__FILE__"shared1");
  r = mod1->SaveByteCode(&stream1);
  if( r < 0 )
    TEST_FAILED;
 
  CBytecodeStream stream2(__FILE__"shared2");
  r = mod2->SaveByteCode(&stream2);
  if( r < 0 )
    TEST_FAILED;
 
  asDWORD crc1 = ComputeCRC32(&stream2.buffer[0], asUINT(stream2.buffer.size()));
 
  r = mod1->LoadByteCode(&stream1);
  if( r < 0 )
    TEST_FAILED;
 
  r = mod2->LoadByteCode(&stream2);
  if( r < 0 )
    TEST_FAILED;
 
  CBytecodeStream stream3(__FILE__"shared1");
  r = mod1->SaveByteCode(&stream3);
  if( r < 0 )
    TEST_FAILED;
 
  CBytecodeStream stream4(__FILE__"shared2");
  r = mod2->SaveByteCode(&stream4);
  if( r < 0 )
    TEST_FAILED;
 
  asDWORD crc2 = ComputeCRC32(&stream4.buffer[0], asUINT(stream4.buffer.size()));
 
  if( crc1 != crc2 )
    TEST_FAILED;
 
  r = mod1->LoadByteCode(&stream3);
  if( r < 0 )
    TEST_FAILED;
 
  r = mod2->LoadByteCode(&stream4);
  if( r < 0 )
    TEST_FAILED;
 
  engine->Release();
}

Unfortunately the test still doesn't reproduce the problem.

Perhaps to reproduce it something specific needs to be registered with the engine too?

Ah, so there was a need to create new engine instances as well.

OK. I'll give it a try tonight.

Yes, now I'm able to reproduce the error. I'll let you know as soon as I have a fix available.

Thanks for the help so far.

I've fixed the problem in revision 1931.

The problem was when loading the bytecode with the funcdef. The funcdef was incorrectly marked as owned by the module. This is what caused the bytecode to be saved slightly differently in the second iteration and ultimately why the bytecode failed to load again.

Regards,

Andreas