Re: [RFA 2/8] Use counted_command_line everywhere

[Tom Tromey <tom@tromey.com>]

>>>>> "Pedro" == Pedro Alves <palves@redhat.com> writes:

>> -	  for (i = 0; i < action->body_count; ++i)
>> -	    trace_dump_actions (action->body_list[i],
>> -				1, stepping_frame, from_tty);
>> +	  trace_dump_actions (action->body_list_0.get (),
>> +			      1, stepping_frame, from_tty);
>> +	  trace_dump_actions (action->body_list_1.get (),
>> +			      1, stepping_frame, from_tty);

Pedro> Hmm, this looked suspicious.  I'm not seeing why would a
Pedro> while-stepping action have two body lists.  I guess this
Pedro> happens to work because trace_dump_actions does nothing
Pedro> if ACTION is NULL.

Yeah.  If you prefer I can add an assert that body_list_1==nullptr,
which seems like it ought to always be true.

>> +      gdb::unique_xmalloc_ptr<char> default_collect_line
>> +	(xstrprintf ("collect %s", default_collect));

Pedro> string_printf ?

I'll do that.

Pedro> It's a shame that we can't use std::make_shared to avoid
Pedro> the separate control block allocation.

It could be done if we changed the representation a bit, but it looked
like a pain to me, so I didn't.

Tom