symbol files in python

symbol files in python

[Dmitry.Neverov via Gdb]

I'm trying to filter out uninitialized variables in a frame by checking
their lines in python, something like this:

sal = frame.find_sal()
frame_line = sal.line
frame_file = sal.symtab.filename

block = frame.block()
while block is not None and block.is_valid():
  for sym in block:
    if sym.line >= frame_line:
      continue
    symtab = sym.symtab
    if symtab is None or symtab.filename != frame_file:
      continue
    # otherwise show symbol as a variable
  if block.function is not None:
      break
  block = block.superblock

Is it correct to expect that symbols in blocks inside a function will
have the same file as the function's frame?

I've noticed that if I add a breakpoint in type::set_num_fields()
(gdb/gdbtypes.h:1002), symtab.filename' for 'this' is 'gdb/mdebugread.c'
and not 'gdb/gdbtypes.h'. Is it expected?

--
Dmitry

Re: symbol files in python

[Tom Tromey]

>>>>> "Dmitry" == Dmitry Neverov via Gdb <gdb@sourceware.org> writes:

Dmitry> I'm trying to filter out uninitialized variables in a frame by checking
Dmitry> their lines in python, something like this:

I suspect this will not yield good results when optimization is present.

I normally see this kind of thing as GIGO -- gdb does what the compiler
tells it.  If the compiler thinks a variable is uninitialized at some
PC, it's free to omit that from the location list.

Dmitry> Is it correct to expect that symbols in blocks inside a function will
Dmitry> have the same file as the function's frame?

Not always, I think because of inlining.  However I'm not really sure.
Compilers can emit pretty much anything they like.

Tom