Re: [Dltk-dev] exception updating variable children count

[Alex Panchenko <alex@xxxxxxxxx>]

Hi Jae,

One of the problems was the parsing of the property children nodes - 
nodeName was not checked.
I have fixed that code.

And another issue is strange numchildren values.

If returns the following context names:

<response xmlns="urn:debugger_protocol_v1" command="context_names" 
transaction_id="12">
<context name="Locals" id="0"/>
<context name="Globals" id="1"/>
*
<context name="Code Objects" id="2"/>*

</response>

We treat 2 as class variables, but here it is "*Code Objects"*, and it 
returns A and B with numchildren=1,
and getObject() function with numchildren=7 (does it mean number of lines?).

At the moment I have added code to fill children that were not retrieved 
from the debugger with instances of the UnknownVariable.

I don't know what these code objects mean for the python developer. 
Probably we should handle them specially, but that should be configured 
for the python only.

Regards,
Alex

Jae Gangemi wrote:
> hello all -
>
>   i keep getting an exception thrown when the code tries to update the 
> children count for python objects. i've tracked down the issue and it 
> seems that for some values returned by the python engine, the children 
> count is less than the available children count. the ScriptValue 
> object's 'variables' field is initialized to be the size of 
> 'property.getChildrenCount', and 'fillVariables' is making a call to 
> getAvailableChildren and an ArrayOutOfBoundsException ends up being 
> thrown b/c the 'variables' array ends up being too small.
>
>   you can see this behavior for yourself by using the attached python 
> script. just set a breakpoint on line 21 (myobject = getObject(1)) and 
> then 'step over'.
>
>   the attached patch fixes this behavior by checking to see if the 
> 'childrenCount' of the dbgp property is less then the number of 
> 'availableChildren' (instead of checking to see if it's less then 0), 
> and if yes, then it sets the 'childrenCount' = availableChildren. as 
> far as i can tell, this change does not seem to negatively impact any 
> of the other debuggers, but perhaps there is a better solution for this.
>
> --
> -jae
> ------------------------------------------------------------------------
>
> _______________________________________________
> dltk-dev mailing list
> dltk-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/dltk-dev