This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

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


Thanks for the review.

On Thursday, December 14 2017, Pedro Alves wrote:

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

Done.

>> +   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...

Removed.

>> +
>> +#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.

Done.

>> +# 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".

Hm, OK.  Added the tests.

>> +/* 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.

Indeed.  I removed it in favour of initializing in-class.

>> --- 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?

Right; endpos is too generic indeed.  Renamed to end_bitpos.

Thanks,

-- 
Sergio
GPG key ID: 237A 54B1 0287 28BF 00EF  31F4 D0EB 7628 65FC 5E36
Please send encrypted e-mail if possible
http://sergiodj.net/


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]