On 6/14/21 10:22 PM, Tom Tromey wrote:
>>>>>> "Tom" == Tom de Vries <tdevries@suse.de> writes:
> 
> Tom> ISTM that the ancestor implementation is confusing and fragile, and should
> Tom> go.  Furthermore, it seems that keeping track of the m_builder in scope can be
> Tom> handled simply with a global variable.
> 
> I'd rather not have any global variables.
> 
> Maybe a new member of dwarf2_per_objfile would be equivalent.
> 

Done.

[ My concern here was problems with inter-obj references, but I've
tested with target board cc-with-dwz-m and didn't run into any problems. ]

> Tom> +  gdb_assert (dwarf2_cu::sym_cu == nullptr);
> Tom> +  scoped_restore restore_sym_cu = make_scoped_restore (&dwarf2_cu::sym_cu);
> Tom> +  dwarf2_cu::sym_cu = cu;
> 
> You can pass 'cu' as a second argument to make_scoped_restore.

Also done, and retested in both test configurations.

Any further comments?

Thanks,
- Tom