Smalltalk › Squeak › Squeak - Dev

debugger experiments

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

4 messages Options

Bob Arning-2

debugger experiments

A small filein to look at a debugger issue. From the preamble:

I'm trying to understand some of the anomalies in what the debugger shows for each step. New versions of two methods in DebuggerMethodMap are added (they can be toggled in and out). Try evaluating

Debugger bobsWay: true; test001
and
Debugger bobsWay: false; test001

and step [Over] until the end of the method. I think the bobsWay: true gives better debugging, but I'm not sure if that holds true everywhere else.

The problem seems to be a slight mismatch in pc's between what the encoder thinks is useful and what the debugger actually stops on, with the additional wrinkle of translating concrete pc's to abstract thrown in.

Cheers,
Bob

debugfixpc.7.cs.gz (3K) Download Attachment

Chris Muller-3

Re: debugger experiments

The difference between the two is, on this line:

1 = 2 ifFalse: [self class].

you have to press Step twice on the class message, which is actually
disconcerting for debugging Smalltalk. bobsWay is definitely how it
should be.

It'd also be really nice if we could fix the "Code Simulation Error"
when stepping Through..

On Mon, Sep 9, 2013 at 6:21 PM, Bob Arning <[hidden email]> wrote:

> A small filein to look at a debugger issue. From the preamble:
>
> I'm trying to understand some of the anomalies in what the debugger shows
> for each step. New versions of two methods in DebuggerMethodMap are added
> (they can be toggled in and out). Try evaluating
>
> Debugger bobsWay: true; test001
> and
> Debugger bobsWay: false; test001
>
> and step [Over] until the end of the method. I think the bobsWay: true gives
> better debugging, but I'm not sure if that holds true everywhere else.
>
> The problem seems to be a slight mismatch in pc's between what the encoder
> thinks is useful and what the debugger actually stops on, with the
> additional wrinkle of translating concrete pc's to abstract thrown in.
>
>
> Cheers,
> Bob
>
>
>

Bob Arning-2

Re: debugger experiments

A little further experimentation.

- dropped some of the invisible nodes related to #to:by:do: to avoid some odd text selections (Debugger test002)
- dropped the abstract pc concept since it doesn't seem to be doing anything useful at the moment

https://dl.dropboxusercontent.com/u/28914094/debugfixpc.9.cs.gz

Cheers,
Bob

On 9/9/13 9:09 PM, Chris Muller wrote:

The difference between the two is, on this line:

    1 = 2 ifFalse: [self class].

you have to press Step twice on the class message, which is actually
disconcerting for debugging Smalltalk.  bobsWay is definitely how it
should be.

It'd also be really nice if we could fix the "Code Simulation Error"
when stepping Through..


On Mon, Sep 9, 2013 at 6:21 PM, Bob Arning [hidden email] wrote:

A small filein to look at a debugger issue. From the preamble:

I'm trying to understand some of the anomalies in what the debugger shows
for each step. New versions of two methods in DebuggerMethodMap are added
(they can be toggled in and out). Try evaluating

Debugger bobsWay: true; test001
and
Debugger bobsWay: false; test001

and step [Over] until the end of the method. I think the bobsWay: true gives
better debugging, but I'm not sure if that holds true everywhere else.

The problem seems to be a slight mismatch in pc's between what the encoder
thinks is useful and what the debugger actually stops on, with the
additional wrinkle of translating concrete pc's to abstract thrown in.


Cheers,
Bob

Bob Arning-2

Re: debugger experiments

In reply to this post by Chris Muller-3

Added an explanation pane to shed some light on mysteries like #to:by:do:. Also an addition to the code pane menu to look at the debugger map.

https://dl.dropboxusercontent.com/u/28914094/debugfixpc.10.cs.gz

Cheers,
Bob

On 9/9/13 9:09 PM, Chris Muller wrote:

The difference between the two is, on this line:

    1 = 2 ifFalse: [self class].

you have to press Step twice on the class message, which is actually
disconcerting for debugging Smalltalk.  bobsWay is definitely how it
should be.

It'd also be really nice if we could fix the "Code Simulation Error"
when stepping Through..


On Mon, Sep 9, 2013 at 6:21 PM, Bob Arning [hidden email] wrote:

A small filein to look at a debugger issue. From the preamble:

I'm trying to understand some of the anomalies in what the debugger shows
for each step. New versions of two methods in DebuggerMethodMap are added
(they can be toggled in and out). Try evaluating

Debugger bobsWay: true; test001
and
Debugger bobsWay: false; test001

and step [Over] until the end of the method. I think the bobsWay: true gives
better debugging, but I'm not sure if that holds true everywhere else.

The problem seems to be a slight mismatch in pc's between what the encoder
thinks is useful and what the debugger actually stops on, with the
additional wrinkle of translating concrete pc's to abstract thrown in.


Cheers,
Bob