Fixed build with GDBJIT enabled and PREJIT disabled

[0xfk0]

This commit fixes build for case, when FEATURE_GDBJIT
is enabled, but FEATURE_PREJIT is disabled.

This solves issue #33956.

[jkotas]

The usage of EEClass::GetSize() in gdbjit.cpp does not make sense. I believe that the right fix should be to replace EEClass::GetSize() class in gdbjit.cpp with sizeof(ArrayBase) or something similar.

[0xfk0]

Looks, like size of the array is complicated thing:

 477     // Set BaseSize to be size of non-data portion of the array
 478     DWORD baseSize = ARRAYBASE_BASESIZE;
 479     if (arrayKind == ELEMENT_TYPE_ARRAY)
 480         baseSize += Rank*sizeof(DWORD)*2;
 481
 482 #if !defined(TARGET_64BIT) && (DATA_ALIGNMENT > 4)
 483     if (dwComponentSize >= DATA_ALIGNMENT)
 484         baseSize = (DWORD)ALIGN_UP(baseSize, DATA_ALIGNMENT);
 485 #endif // !defined(TARGET_64BIT) && (DATA_ALIGNMENT > 4)
 486     pMT->SetBaseSize(baseSize);

May be it's better to use GetBaseSize() method, like this:

  m_type_size = typeHandle.AsMethodTable()->GetBaseSize();

Unfortunaty, I not understood, what is the difference between MethodTable::GetBaseSize and EEClass::GetSize

But practically, looks like this information (size of data type) isn't so critical -- it only be written to DWARF, and I see no sense, why debugger should want to examine array object's contents (may be I missing something?)

Situation looks like following:

1. GetTypeInfoFromTypeHandle for arrays returns NamedRefTypeInfo class containing ClassTypeInfo class instance (in constructor of which GetSize() was previously used);

2. m_type_size variable is used only in implementations of DumpDebugInfo
   function;

3. DumpDebugInfo function is separately implemented in each class
   inherited from TypeInfoBase and DwardDumpable, so ClassTypeInfo have
   own implementation;

4. m_type_size used in ClassTypeInfo::DumpDebugInfo to write out portion of DWARF file, describing particular data type, m_type_size specifies the size of the type (it isn't used to read from memory, etc...)

So, from one point of view it's safe (at least nothing crashes) to replace existing code (call to
MethodTable::GetBaseSize instead of EEClass::GetSize).

From other side, isn't clear why originally EEClass::GetSize was needed.

I don't know what to do. I will update commit...

[0xfk0]

@gbalykov @alpencolt @y-yamshchikov

[jkotas]

No compiled into default CoreCLR builds - no need to wait for CI to finish.