[Lldb-commits] [lldb] r255525 - Make debug info specification use categories system.

[Zachary Turner via lldb-commits]

Yea I realized this after.  At this point everything seems to be working
and I need to finish other stuff, but I might come back to this if I have
spare cycles.  Anyone else feel free to do the same.

On Tue, Dec 15, 2015 at 3:16 AM Tamas Berghammer <tberghammer at google.com>
wrote:

> I think the problem is that you don't make a copy of the category variable
> manually. The function you are creating capture category by reference and
> when you update the value of category in the next loop iteration you also
> update the value captured by the previous function. As a result all of your
> test method will have the same category value. The canonical way to solve
> these issues is to add a new argument called category and set its default
> value to category (same as I did for argvalue).
>
> On Mon, Dec 14, 2015 at 11:01 PM Zachary Turner via lldb-commits <
> lldb-commits at lists.llvm.org> wrote:
>
>> Alright, I'll revert this loop for now.  This is in at r255567.
>>
>> On Mon, Dec 14, 2015 at 2:33 PM Siva Chandra <sivachandra at google.com>
>> wrote:
>>
>>> On Mon, Dec 14, 2015 at 2:15 PM, Zachary Turner <zturner at google.com>
>>> wrote:
>>> > Diff looks like this:
>>> >
>>> > -                @dwarf_test
>>> > -                @wraps(attrvalue)
>>> > -                def dwarf_test_method(self, attrvalue=attrvalue):
>>> > -                    self.debug_info = "dwarf"
>>> > -                    return attrvalue(self)
>>> > -                dwarf_method_name = attrname + "_dwarf"
>>> > -                dwarf_test_method.__name__ = dwarf_method_name
>>> > -                newattrs[dwarf_method_name] = dwarf_test_method
>>> >
>>> > ...
>>> >
>>> > +                for category in supported_categories:
>>> > +                    @add_test_categories([category])
>>> > +                    @wraps(attrvalue)
>>> > +                    def test_method(self, attrvalue=attrvalue):
>>> > +                        self.debug_info = category
>>> > +                        return attrvalue(self)
>>> > +                    method_name = attrname + "_" + category
>>> > +                    test_method.__name__ = method_name
>>> > +                    newattrs[method_name] = test_method
>>> >
>>> >
>>> > So it looks like it's still under a different function.  The
>>> difference is
>>> > that before (in the - section of the diff) there were 3 different
>>> functions
>>> > with 3 different names.  For example `dwarf_test_method` above.  In
>>> the new
>>> > code (+ section of the diff) it looks like one function defined 3
>>> different
>>> > times with the same name `test_method`.  But each of these should be a
>>> > distinct function object that is defined in the context of the closure
>>> > determined by the enclosing scope (the for loop in this case).  So all
>>> 3
>>> > functions are actually different.  self is just an argument to the
>>> function,
>>> > not some magic value like `this` in C++, so as in the previous
>>> version, the
>>> > `debug_info` that gets modified is on whatever instance of `self` gets
>>> > passed into the function, same as before.
>>>
>>> I think you are right. I am only basing my theories by feeling the
>>> elephant blind here. When running TestWithLimitDebugInfo, I have
>>> printed the |category| value from test_method and it only prints
>>> "dwo". So, it is |category| which is at fault here as it is the loop
>>> variable. Though two different methods are actually getting added,
>>> self.debug_info refers to the last value of |category|.
>>>
>>> > Tamas or Pavel can probably clarify, but that's my understanding.
>>>
>> _______________________________________________
>
>
>> lldb-commits mailing list
>> lldb-commits at lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/lldb-commits/attachments/20151215/8752b46a/attachment.html>