This is the mail archive of the gdb-patches@sources.redhat.com 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: Start Fortran support for variable objects.


 > I have two comments below, hope that they might be helpful.  Thanks.
 > 
 > On Fri, 1 Jul 2005, Nick Roberts wrote:
 > 
 > > As in the patch below?  I don't understand the extra cases it appears to
 > > cover, but it worked for the tests I tried.
 > 
 > Maybe it is also helpful to test against such arrays definition as:
 >   
 >   integer array(0:5), integer array(-1:4)
 > 
 > or even
 >   
 >   integer array(0:5,-1:4)
 >
 > (if variable object does support multi-dimension array. I don't know much 
 > about variable object, and MI as a whole.)

Ah!  I see now.  I've not used such arrays in Fortran.  Using

      INTEGER ARRAY1(0:5), ARRAY2(-1:4)
      INTEGER ARRAY3(0:2,-1:1)
      DATA ARRAY1/1,2,3,4,5,6/
      DATA ARRAY2/1,2,3,4,5,6/
      DATA ARRAY3/11,21,31,12,22,32,13,23,33/

the latest patch seems to work (see attached image below).  I am sure
that my original patch would have failed for these cases.

 > The second comment is about the following text in a former mail Nick sent: 
 > 
 > > Fortran:
 > > 
 > > (top-gdb) p TYPE_LOW_BOUND(var->type)
 > > $3 = 0
 ...
 > 
 > For Fortran array such as DIMENSION I(4), the lower bound should be 1 by
 > default.  The following session on my box shows this:
 > 
 > <top-gdb> p type->main_type->fields->type->main_type->fields[0].loc.bitpos
 > $3 = 1
 > <top-gdb> p type->main_type->fields->type->main_type->fields[1].loc.bitpos
 > $4 = 4

So I should have done:

(top-gdb) p TYPE_LOW_BOUND(var->type->main_type->fields->type)
$1 = 1
(top-gdb) p TYPE_HIGH_BOUND(var->type->main_type->fields->type)
$2 = 4

 > I guess there might be some errors in the process of creating varobj for 
 > Fortran array.  

No the information seems to be there.  So maybe:

  for (i = 0; i < var->num_children; i++)
    {
      /* Mark as the end in case we bail out */
      *((*childlist) + i) = NULL;

      j = i + TYPE_LOW_BOUND(var->type->main_type->fields->type);

      /* check if child exists, if not create */
      name = name_of_child (var, j);
      child = child_exists (var, name);
      if (child == NULL)
	child = create_child (var, j, name);

      *((*childlist) + i) = child;
    }

will work in varobj_list_children in a language independent way.

I'll wait to see what Daniel says though, before submitting another patch.

Nick


PNG image


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