[Gambas-user] a Debug question

[Benoît Minisini]

> Benoît Minisini ha scritto:
> >> If  one puts in a breakpoint, and is  stepping through the code one line
> >> at a time to check values of variables etc, is there any way or
> >> re-setting to a line in the same subroutine prior to the current line
> >> you have stopped on.
> >>
> >> I've found sometimes I'm trying to find a bug, and I put in multiple
> >> break points so that I can skip from one to another, and then want to
> >> check something in the middle that I've gone past, but have to stop the
> >> entire program and start again.
> >
> > Alas you cannot go back in time in debuggers yet. I've heard about people
> > thinking about it, but I don't think you will never see it in Gambas!
>
> This is a serious problem, tough. There are programs which, for one
> reason or another, are a little difficult to debug using the internal
> gambas debugger. For example, think at a data-entry form, with a lot of
> controls, each having validation (perhaps, put on top of a database). If
> you want to debug the final action of the form (the Apply button, may
> be), then you have to fill in all the values. Every time you discover a
> bug, even a stupid one, you must restart the program from the beginning;
> and, may be, that form is nested deep in the program logic - you have to
> step through many menus and dialogs to arrive at the point of interest.
> Other times, you don't have a clear idea of the algorithm you have to
> write, so you try again and again. In these situations, the more you can
> do through the debugger, the more time you save.
>
> There could be some improvements to the debugger:
>
>     - the possibility to assign new values to variables (perhaps not the
> structured, but only the simplest ones)

This is planned, but some function is yet missing in the gb.eval command to 
implement ot.

>     - the possibility to modify the program text and continue debugging
> (perhaps without adding/deleting lines)

I think I can't do that. This is too much work for something that is not 
really useful.

>     - the possibility to change the program counter (perhaps with
> constraints)

I don't find that useful at all, but I think it may be possible to implement.

>     - the possibility to add "debug" code, which is compiled in only
> when a special "debug run" is invoked

This is planned.

>     - conditional breakpoints / break on the "Nth occurence"
>

This is not planned, but it is possible to implement. I will note that.

> ...and may be other which does'nt come to my mind at the moment.
>
> We already discussed a little about this: I know this is somehow hard to
> implement, but having the gambas virtual machine ready in memory should
> help... I mean: if the debugger "knows" how to obtain the value of a
> variable, and how to display it, then perhaps it is possible to
> [relatively] easily assign a new value. Or, the possibility to modify
> the source and continue debugging, seems to me not very hard to do (but
> I could be wrong, of course).
>
> I know that I am biased, because for my work I go deep to the processor
> hardware, and the debugger lets me do nearly everything I want, but you
> could consider whether any improvement could be done to the gambas
> debugger.
>
> Hope I did'nt annoyed you, they were simply ideas.

You didn't annoyed me. :-)

Regards,

-- 
Benoît