Re: [hatari-devel] debugger help

[Eero Tamminen <oak@xxxxxxxxxxxxxx>]

Hi,

On perjantai 21 helmikuu 2014, Douglas Little wrote:
> Ok I might be speaking too soon here.. but..
> 
> ...yes that is already awesome :-) A quick test over lunch with the patch
> applied proves that I can now trace through code in logical PC order
> using 'n' command alone, without having to figure out what each branch
> is going to do, or if it's a control-flow op like RTS.

Known issues:
- Things that "n" doesn't skip yet: ftrapcc, chk2
- Things that "n branch" doesn't match: fbcc, fdbcc

If there's any other issue, I would be interested to know.


> Tracing just got 10+ times easier :) The pain is gone!

Ok, I'll commmit the patch and document it to manual.


> NB I also didn't notice any interrupts getting in the way during tracing,
> although I was running until a breakpoint rather than breaking in
> forcibly. I'll report on this separately after I've had some time to
> play with it.

I didn't do anything for that.  "n" will skip exception
instructions like "trap", but that's all.


	- Eero

> 
> 
> Thanks Eero!
> 
> D
> 
> On 20 February 2014 21:23, Eero Tamminen <oak@xxxxxxxxxxxxxx> wrote:
> > Hi,
> > 
> > On torstai 13 helmikuu 2014, Douglas Little wrote:
> > > I will give that a try, thanks! :-)
> > 
> > Does it work OK now?
> > 
> > (I'm wondering should I commit it to repo & document in manual.)
> > 
> >         - Eero
> > > 
> > > Doug.
> > > 
> > > On 12 February 2014 21:05, Eero Tamminen <oak@xxxxxxxxxxxxxx> wrote:
> > > > Hi,
> > > > 
> > > > On tiistai 11 helmikuu 2014, Douglas Little wrote:
> > > > > If "next" should behave like "next" in other debuggers, it would
> > > > > 
> > > > > > need to know what the current instruction will do, and
> > > > > > depending on that, either set breakpoint like now, or step one
> > > > > > instruction.
> > > > > > 
> > > > > > Question is, what kind of instructions would require breakpoint
> > > > > > instead of stepping: BSR, JSR, BKPT, ILLG, STOP, TRAP...?
> > > > > 
> > > > > I agree this is probably difficult/complicated to implement in
> > > > > terms of predicting a breakpoint address. It's probably more
> > > > > comfortable as a hook in the emulation stepper itself, which
> > > > > informs the debugger of each new PC visited and provides an
> > > > > opportunity to break. Something like that which doesn't require
> > > > > the debugger to understand a lot of detail on what happened -
> > > > > shoves the work onto the emulator.
> > > > > 
> > > > > I don't know if that is something that fits well with Hatari
> > > > > and/or the existing cores - or what other side effects this
> > > > > would have on performance, if it could even be made optional
> > > > > etc. so that's another matter entirely...
> > > > 
> > > > Does the attached patch do what you were after?
> > > > 
> > > > (*_OpcodeType() functions have some deficiencies in them, but
> > > > those can be improved separately.)
> > > > 
> > > >         - Eero