[PATCH v6 2/2] Implement pahole-like 'ptype /o' option

Pedro Alves palves@redhat.com
Thu Dec 14 14:50:00 GMT 2017


On 12/14/2017 02:48 AM, Sergio Durigan Junior wrote:

> A big part of this patch handles the formatting logic of 'ptype',
> which is a bit messy.  The code to handle bitfield offsets, however,
> took some time to craft.  My thanks to Pedro Alves for figuring things
> out and pointing me to the right direction, as well as coming up with
> a way to inspect the layout of structs with bitfields (see testcase
> for comments).

Why, thanks. :-)

> +Issuing a @kbd{ptype /o struct tuv} command would print:
> +
> +@smallexample
> +(@value{GDBP}) ptype /o struct tuv
> +/* offset    |  size */
> +struct tuv @{
> +/*    0      |     4 */    int a1;
> +/* XXX  4-byte hole  */
> +/*    8      |     8 */    char *a2;
> +/*   16      |     4 */    int a3;
> +@} /* total size:   24 bytes */
> +@end smallexample

These examples need to be updated per the new output format.

> +   You should have received a copy of the GNU General Public License
> +   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
> +
> +/* This file will be used to test 'ptype /o' on x86_64 only.  */

No longer true...

> +
> +#include <stdint.h>
> +

> +
> +# Test only works on x86_64 LP64 targets.  That's how we guarantee

Remove reference to x86_64; it's no longer true.

> +# that the expected holes will be present in the struct.
> +if { ![is_lp64_target] } {
> +    untested "test work only on lp64 targets"
> +    return 0
> +}
> +
> +if { [prepare_for_testing "failed to prepare" $testfile $srcfile \
> +	  { debug c++ }] } {
> +    return -1
> +}

I think we're missing a test for "ptype /oTM", to make sure that
we can still override the fact that /o implies /tm ?
I'd add at least one for "/oTM" and one for "/TMo".  The
latter ends up being the same as "/o".

> +/* The default values for a struct print_offset_data.  */
> +
> +struct print_offset_data print_offset_default_data =
> +{
> +  0,				/* offset_bitpos */
> +  0,				/* endpos */
> +};
> +

Do we really need this object ?  How about just defining
the default values in-class ?  See below.

> --- a/gdb/typeprint.h
> +++ b/gdb/typeprint.h
> @@ -24,6 +24,22 @@ struct ui_file;
>  struct typedef_hash_table;
>  struct ext_lang_type_printers;
>  
> +struct print_offset_data
> +{
> +  /* The offset to be applied to bitpos when PRINT_OFFSETS is true.
> +     This is needed for when we are printing nested structs and want
> +     to make sure that the printed offset for each field carries over
> +     the offset of the outter struct.  */
> +  unsigned int offset_bitpos;

So here:

  unsigned int offset_bitpos = 0;

> +
> +  /* ENDPOS is the one-past-the-end bit position of the previous field
> +     (where we expect the current field to be if there is no
> +     hole).  */
> +  unsigned int endpos;

and

   unsigned int endpos = 0;

Then you default-constructed print_offset_data objects have
the fields automatically zeroed:

  print_offset_data podata;

while at it, wouldn't it be better to name this one "end_bitpos" or
something like that with "bit" in it as well?

Thanks,
Pedro Alves



More information about the Gdb-patches mailing list