Re: [PATCH v2 4/5] [gdb/symtab] Improve invalid range check in create_addrmap_from_gdb_index

[Tom de Vries <tdevries@suse.de>]

On 8/21/25 15:31, Tom de Vries wrote:
> +      if (valid_range_p)
> +	{
> +	  CORE_ADDR relocated_lo
> +	    = per_objfile->relocate (unrelocated_addr (lo));
> +	  CORE_ADDR relocated_hi_m1
> +	    = per_objfile->relocate (unrelocated_addr (hi_m1));
> +	  struct obj_section *lo_sect = find_pc_section (relocated_lo);
> +	  struct obj_section *hi_sect = find_pc_section (relocated_hi_m1);
> +	  valid_range_p = lo_sect != nullptr && hi_sect != nullptr;
> +	}

It turns out that this doesn't always work.

I ran the test-case gdb.base/code_elim.exp (which uses add-symbol-file) 
with target board cc-with-gdb-index, and ran into the following problem.

Function addr_info_make_relative gets called, which puts the section 
table into some relative address state.  Consequently, the relocated 
address are also relative, and aren't found using find_pc_section.

I'll submit a v3 once I've fixed this.

Thanks,
- Tom