diff mbox

Adding bit position to ptype

Message ID 20140618031308.GA7252@ubuntu
State New
Headers show

Commit Message

lin zuojian June 18, 2014, 3:13 a.m. UTC
Hi, I am having trouble looking up field offsets when reading disassembled text.
So I add an extra field to ptype:
(gdb) ptype WebCore::FrameLoader
type = class WebCore::FrameLoader {
  private:
    WebCore::Frame *m_frame bitpos: 0;
    WebCore::FrameLoaderClient *m_client bitpos: 32;
    WebCore::PolicyChecker m_policyChecker bitpos: 64;
    WebCore::HistoryController m_history bitpos: 3264;
    WebCore::ResourceLoadNotifier m_notifer bitpos: 3552;
    WebCore::SubframeLoader m_subframeLoader bitpos: 3584;
    WebCore::FrameLoaderStateMachine m_stateMachine bitpos: 3680;
    WebCore::IconController m_icon bitpos: 3712;
...

That helps me a lot to understand what "ldr r0, [r4, #408]" means.
---
 gdb/c-typeprint.c | 6 ++++++
 1 file changed, 6 insertions(+)

Comments

Tom Tromey June 18, 2014, 2:31 p.m. UTC | #1
>>>>> ">" == lin zuojian <manjian2006@gmail.com> writes:

>> Hi, I am having trouble looking up field offsets when reading
>> disassembled text.  So I add an extra field to ptype:

Nice.

You might be interested in pahole.py.

>> diff --git a/gdb/c-typeprint.c b/gdb/c-typeprint.c
>> index 4edc9ec..9fbbc36 100644
>> --- a/gdb/c-typeprint.c
>> +++ b/gdb/c-typeprint.c
>> @@ -1110,6 +1110,12 @@ c_type_print_base (struct type *type, struct ui_file *stream,
>>  		    fprintf_filtered (stream, " : %d",
>>  				      TYPE_FIELD_BITSIZE (type, i));
>>  		  }
>> +		else if (!field_is_static (&TYPE_FIELD (type, i)))
>> +		  {
>> +		    /* print the bit pos.  */
>> +		    fprintf_filtered (stream, " bitpos: %d",
>> +		                      TYPE_FIELD_BITPOS (type, i));
>> +		  }
>>  		fprintf_filtered (stream, ";\n");

To be considered for acceptance all patches require a ChangeLog entry
and must pass regression testing -- which this one probably does not, as
it changes gdb's output by default.  Please see the contribution
instructions.

I think it would probably be better to add a flag to ptype to enable
this behavior.  Integrating all of the pahole behavior into ptype would
be even better, though I understand if you don't want to go this far.

thanks,
Tom
lin zuojian July 1, 2014, 9:33 a.m. UTC | #2
Hi Tom,
    Before posting the second patch, I may make your point clear first.
    Do you want me to add a flag to control this behavior?
    Thanks
--
Lin Zuojian
Tom Tromey July 1, 2014, 4:29 p.m. UTC | #3
>>>>> ">" == lin zuojian <manjian2006@gmail.com> writes:

>>     Before posting the second patch, I may make your point clear first.
>>     Do you want me to add a flag to control this behavior?

Yeah, I think the behavior should be optional and should default to
"off".

I think it's best to consider this as a new feature -- so it would need
a new ptype flag, documentation, NEWS entry, and tests.  The bad news is
that this is much more work for you.  That's probably why nobody has
implemented it before.

Tom
diff mbox

Patch

diff --git a/gdb/c-typeprint.c b/gdb/c-typeprint.c
index 4edc9ec..9fbbc36 100644
--- a/gdb/c-typeprint.c
+++ b/gdb/c-typeprint.c
@@ -1110,6 +1110,12 @@  c_type_print_base (struct type *type, struct ui_file *stream,
 		    fprintf_filtered (stream, " : %d",
 				      TYPE_FIELD_BITSIZE (type, i));
 		  }
+		else if (!field_is_static (&TYPE_FIELD (type, i)))
+		  {
+		    /* print the bit pos.  */
+		    fprintf_filtered (stream, " bitpos: %d",
+		                      TYPE_FIELD_BITPOS (type, i));
+		  }
 		fprintf_filtered (stream, ";\n");
 	      }