Hi,
This patch series proposes a feature that GDB is able to visit varobj
available children only, by adding a new option
"--available-children-only" to commands -var-create,
-var-info-num-children and -var-list-children. This patch series was
written by Pedro in one patch. I read the patch three months ago, split
it, add some minor things, and write test cases. I wish I didn't
break anything, including the rationale and the implementation :)
In each traceframe, with option --available-children-only, the
children of varobj vary, so it behaves like a dynamic varobj.
Since the key of dynamic varobj is the iterator, which can be used to
visit each child. We need a new iterator for varobj with
--available-children-only.
Current dynamic varobj is python pretty-printer specific, so the
iterator is python specific too. In order to add a new type of iterator,
we have to generalize iterator and de-couple it from python. At
present, iteration is performed against PyObject, but it can
generalized to a name-value pair. That is what patch #1 ~ #3 do.
Note that patch #2 and #3 can be a single commit, but I split it for
review. Ideally, dynamic varobj can be a generic stuff, IMO.
After the changes in patch #1 ~ #3, we find some code are not
python-specific, so #if HAVE_PYTHON can be removed. It is done by
patch #4.
Since we think dynamic varobj is no longer python specific, we'd
better rename predicate varobj_pretty_printed_p to varobj_is_dynamic_p,
and use it more widely (done by patch #5 #6). It paves a way for
the next patches to add a new type of dynamic varobj.
Patch #7 adds the option --available-children-only in MI code, patch
#8 adds the corresponding iterator, and patch #9 updates varobj children
when traceframe is changed. These three should be in one commit.
Note that there is at most one iterator can be applied to one varobj.
If pretty-printer is installed and option --available-children-only
is used, GDB has to return one iterator, and we choose iterator
for available-children-only.
Patch #10 and #11 are about test cases. Patch #10 is to teach
testsuite to match dynamic="1", and patch #11 is the test case for
option --available-children-only.
I don't include NEWS and doc patch in this series, because we'd like
to discuss on "whether we can call varobj with --available-children-only
a dynamic varobj". The result affects the doc, IMO. Although we
implemented varobj with --available-children-only as a dynamic varobj,
I am not sure we can do the same in doc, which is user visible.
The whole series is tested on x86_64-linux.