Re: [PATCH v2 4/4] gdb/dwarf: use dynamic partitioning for DWARF CU indexing

[Simon Marchi <simon.marchi@efficios.com>]

On 9/18/25 3:26 PM, Tom Tromey wrote:
>>>>>> "Simon" == simon marchi <simon.marchi@polymtl.ca> writes:
> 
> Simon> +  /* Launch parallel tasks to index units.  */
> Simon> +  gdb::parallel_for_each_async<1, dwarf2_per_cu_up *, parallel_indexing_worker>
> Simon> +    (per_bfd->all_units.data (),
> Simon> +     per_bfd->all_units.data () + per_bfd->all_units.size (),
> Simon> +     [this] () { this->done_reading (); }, "DWARF indexing worker", this);
> Simon>  }
>  
> I was curious to know why this uses .data() rather than iterators.
> 
> Tom

Yeah, that's unfortunate.  It's because
std::atomic<std::vector<...>::iterator> won't work with
-D_GLIBCXX_DEBUG.  If I try to use iterators all the way, I get:

/opt/gcc/15.2.0/include/c++/15.2.0/atomic:214:21: error: static assertion failed: std::atomic requires a trivially copyable type
  214 |       static_assert(__is_trivially_copyable(_Tp),
      |                     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~

Without _GLIBCXX_DEBUG, vector iterators boil down to just a pointer.
But with _GLIBCXX_DEBUG, there is additional state.

I still want to be able to use _GLIBCXX_DEBUG, it's pretty useful.

Simon