Ghidra dependency resolve can report incorrect recursive type deps

[mahaloz]

This issue is related to #97, since the root bug is scopes in Ghidra.

This the fdupes binary inside the CI, the following can be observed:

[ins] In [10]: func = deci.functions[7526]

[ins] In [11]: func.args
Out[11]: {0: <FuncArg: char * filename; @0>, 1: <FuncArg: __off64_t max_read; @1>}

[ins] In [12]: for name, t in deci.typedefs.items():
          ...:     if "__off64_t" in name:
          ...:         print(t)
          ...: 
<TypeDef: DWARF/__off64_t=sqword>
<TypeDef: types.h/__off64_t=long>

[ins] In [13]: deci.get_dependencies(func)
Out[13]: 
[<TypeDef: DWARF/md5_word_t=uint>,
 <TypeDef: DWARF/__off64_t=sqword>,
 <Struct: DWARF/md5.h/md5_state_s membs=3 (0x58)>,
 <TypeDef: DWARF/md5_byte_t=uchar>,
 <TypeDef: DWARF/md5.h/md5_state_t=md5_state_s>,
 <GlobalVar: md5_byte_t[16] digest; @2122960 len=16>,
 <GlobalVar: md5_byte_t[8192] chunk; @2122976 len=8192>]

The issue here is the second arg of the function has type __off64_t, and there are two typedefs in Ghidra that define this type. Unfortunately, we choose the wrong one for this reference. We chose DWARF, but we should've chosen types.h. In #146 a temporary solution was chosen always to prioritize types.h, but this will not work into the future.

To solve this, we must encode scopes into the artifacts, which will require more time to develop.