Re: [PATCH] gdb.dwarf2: Define and use gdb_target_symbol_prefix for symbol prefixes

[Pedro Alves <palves@redhat.com>]

On 10/30/2015 04:25 AM, Kevin Buettner wrote:

> diff --git a/gdb/testsuite/lib/gdb.exp b/gdb/testsuite/lib/gdb.exp
> index 048070b..f8e35ac 100644
> --- a/gdb/testsuite/lib/gdb.exp
> +++ b/gdb/testsuite/lib/gdb.exp
> @@ -5521,6 +5521,45 @@ proc core_find {binfile {deletefiles {}} {arg ""}} {
>      return $destcore
>  }
>  
> +# gdb_target_symbol_prefix compiles a test program and uses readelf
> +# to determine the prefix (such as underscore) for linker symbol
> +# prefixes.
> +
> +proc gdb_target_symbol_prefix {} {

Shouldn't this be gdb_caching_proc ?

Then T wonder whether something like:

-                    DW_OP_addr table_2_ptr
+                    DW_OP_addr [gdb_symbol table_2_ptr]

instead of prepending the $prefix variable results in clearer
test code.

> +	set readelf_program [gdb_find_readelf]
> +	set result [catch "exec $readelf_program --syms $exe" output]
> +
> +	if { $result == 0 \
> +	     && ![regexp { ([^ a-zA-Z0-9]*)main$} $output dummy prefix] } {
> +	    verbose "gdb_target_symbol_prefix: Could not find main in readelf output; returning null prefix" 2
> +	}

I'm wondering about using a method that would work for mingw/cygwin
as well.  The only existing use of the target symbol prefix in the
tree is for those targets, which are coff/pe, not elf.  See
gdb_target_symbol_prefix_flags.  Maybe just use objdump instead?

Thanks,
Pedro Alves