Re: [Simulavr-devel] avrtest and simulavr vs simulavrxx

[Klaus Rudolph]

Joerg Wunsch schrieb:
> As Klaus Rudolph wrote:
>
> > > -C option: provide a "coredump" after simulation finished
>
> > Is there a typical format for this?
>
> It's a fairly homegrown (of the old simulavr) text file.  I'm
> attaching one for reference.  Also, it's always using a fixed name.
>
> Changing the format might be an option, though the human readable text
> file for sure does have some merit.  Right now, avr-libc's testsuite
> uses a Python script to extract the desired information (exitcode from
> r24/r25, and possibly a message string from "external memory" starting
> at address 0x2000).  This allows for a fairly simple interface to pass
> a result code and possibly a message from the testsuite C source file
> to the user.
objdump will do the job as well.

> > Because it is mostly a memory dump it could be binary from address 0 to 
> > end_of_ram/flash or it could be a complete elf-file which also includes 
> > the debug symbols from the loaded elf prog.
>
> Host systems typically use ELF as the coredump format, yes, but they
> don't duplicate the information that was already present in the actual
> executable file.  Thus, when using a debugger to analyze a coredump
> file, you always have to pass both, the executable as well as the dump
> itself to the debugger.

Because avr devices have a very small memory footprint it is no problem 
to duplicate the programm/symbols content to the core file. This has the 
benefit that you can run a lot simulations with different modifications 
to your source and don't have to remember which pair of core file and 
executable match.

> > > The avr-libc testsuite uses the option -B <addr>, aka. --breakpoint
> > > <addr>, in order to stop the simulation.  [...]
>
> > I can add that. Is there a new option format required like 0x... 0000h 
> > or something? I have no idea which format simulavrxx already have :-)
>
> Thanks for adding it.  The format is a standard C string so you could
> use strtoul(..., 0).
Actually we use strtol(,16).


And I found a nice "feature"!

I have a function named f1(). If you add -Tf1 you will have a breakpoint 
at 0xf1! :-) Not really the expected behavior :-) Maybe we should add 
this to the doc? Should we do a check if symbols matches against params 
also if numbers found? Or should we do both and give a warning?