Re: [PATCH 5/6] Return void from buildsym_compunit::push_context

[Tom Tromey <tromey@adacore.com>]

>>>>> "Simon" == Simon Marchi <simark@simark.ca> writes:

Simon> On 4/17/26 2:24 PM, Tom Tromey wrote:
>> diff --git a/gdb/dwarf2/read.c b/gdb/dwarf2/read.c
>> index 4447923ae3e..2a330655c48 100644
>> --- a/gdb/dwarf2/read.c
>> +++ b/gdb/dwarf2/read.c
>> @@ -7799,10 +7799,11 @@ read_func_scope (struct die_info *die, struct dwarf2_cu *cu)
>> }
>> }
>> 
>> -  gdb_assert (cu->get_builder () != nullptr);
>> -  context_stack &ctx = cu->get_builder ()->push_context (lowpc);
>> +  buildsym_compunit *builder = cu->get_builder ();
>> +  gdb_assert (builder != nullptr);
>> +  builder->push_context (lowpc);
>> symbol *func_sym = new_symbol (die, read_type_die (die, cu), cu, templ_func);
>> -  ctx.name = func_sym;
>> +  builder->set_current_context_function (func_sym);

Simon> It might be tricky depending on how new_symbol uses the current context
Simon> when reading a DW_TAG_subprogram, but I wonder if we could call
Simon> new_symbol first, then pass the resulting symbol to push_context.  Then
Simon> we wouldn't need to have a setter.  If that's possible, I think it would
Simon> be slightly better, because it would show that the symbol is initialized
Simon> when the context is pushed and can't change.  Other callers of
Simon> push_context would pass a nullptr symbol.

I had looked at this when writing the patch but I concluded that this
wasn't possible because new_symbol calls add_symbol_to_list.

Ideally new_symbol would be removed and everything cleaned up.  But
that's difficult.

Tom