Re: [PATCH 1/1] gdb: Fix assertion in 'value_primitive_field'

[Tom Tromey <tom@tromey.com>]

>>>>> "Stephan" == Stephan Rohr <stephan.rohr@intel.com> writes:

Stephan> From: "Rohr, Stephan" <stephan.rohr@intel.com>
Stephan> GDB asserts that the data location of a value's field with a dynamic
Stephan> data location is resolved when fetching the field's value in
Stephan> 'value_primitive_field'.  This assertion was hit because of bogus DWARF
Stephan> generated by the Intel Fortran compiler.  While that compiler should fix
Stephan> the DWARF, we prefer GDB to error out here instead, e.g. to allow the
Stephan> user to continue debugging other parts of the program.

It's hard to be sure in this code, but normally if there's an assert, it
means that there's a requirement that the caller do something.  So maybe
this is a bug somewhere else?

The code does seem self-contradictory though:

      /* We expect an already resolved data location.  */
      gdb_assert (TYPE_DATA_LOCATION (type)->is_constant ());
      /* For dynamic data types defer memory allocation
	 until we actual access the value.  */
      v = value::allocate_lazy (type);

Like how would it be resolved but also possibly dynamic.

But since this is allocating a lazy value, why does is_constant even matter?
What happens if you just remove that assert entirely?

I think some kind of test case here would be good.

Tom