Ghidra assigns stack address to extra variable

[cgrzemba]

I have a decompilation result, where I get a variable (puVar12) with a stack address:

puVar12 = &stack0xffffff60;
lVar19 = __div64(CONCAT44((iVar6 - (int)((ulonglong)lVar19 >> 0x20)) -
                              (uint)(uVar13 < (uint)lVar19),uVar13 - (uint)lVar19),1000000);

The signature of function locks good

longlong __div64 (longlong param_1, longlong param_2)

Why do Ghiidra add this extra variable?

An other strange thing is the 'CONCAT44 and >> 0x20' which is not related to any assembler code. It seems Ghidra tries to handle a compiler specific data type. lVar19 contains a hrtime_t value.
Where it is come from?

0x0805af51 <+69>:    mov    esi,DWORD PTR [ebp+0x98]
   0x0805af57 <+75>:    mov    ebx,DWORD PTR [ebp+0x9c]
   0x0805af5d <+81>:    mov    ecx,esi
   0x0805af5f <+83>:    sub    ecx,eax
   0x0805af61 <+85>:    mov    eax,ebx
   0x0805af63 <+87>:    sbb    eax,edx
   0x0805af65 <+89>:    push   0x0 
   0x0805af67 <+91>:    push   0xf4240
   0x0805af6c <+96>:    push   eax 
   0x0805af6d <+97>:    push   ecx 
   0x0805af6e <+98>:    call   0x80570e4 <__div64@plt>
   0x0805af73 <+103>:   lea    ecx,[edi+0x1]
   0x0805af76 <+106>:   cmp    eax,ecx

It is a Sun Studio Compiled code for Solaris.

In the C code listing all functions are mared with "Unknown calling convention". It is possible to assign there "__cdecl" for all functions globally?

Many Thanks!

[cgrzemba]

duplicate