Re: [PATCH] gdb/dwarf2: don't search included symtabs recursively

[Simon Marchi <simark@simark.ca>]

On 2/5/26 1:53 PM, Tom Tromey wrote:
>>>>>> "Simon" == Simon Marchi <simon.marchi@efficios.com> writes:
> 
> Simon> It therefore seems unnecessary to do a recursive search, in
> Simon> recursively_find_pc_sect_compunit_symtab, it will search some symtabs
> Simon> multiple times.
> 
> Agreed, thanks for finding this.
> 
> We discussed this elsewhere but probably what we eventually want is a
> mechanism to avoid searching included symtabs more than once across an
> entire search.
> 
> Simon> I am not sure how blockvectors work exactly, whether the blockvector of
> Simon> the includer CU is a superset of the blockvectors of the includees.  I
> Simon> ask that because it seems like in practice, the requested PC always
> Simon> seems to be found in the first searched CU.  I haven't investigated this
> Simon> point more than that though.
> 
> I think each compunit_symtab gets a blockvector that describes exactly
> the functions defined locally in that CU.
> 
> It would be very unusual for a partial unit to really contain code.  The
> reason for this is that the normal reason to make a partial unit is to
> share common things between CUs, but two CUs would not normally describe
> the same function with the same address ranges.  It wouldn't be invalid
> to do this, just not very useful.
> 
> Approved-By: Tom Tromey <tom@tromey.com>
> 
> Tom

Ok, so this "problem" wasn't really a problem in practice, because a PC
would always be found in the first (top-level) compunit_symtab then
(unless it's not found at all).

Pushed, thanks.

Simon