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

[Klaus Rudolph]

> -C option: provide a "coredump" after simulation finished
>
> The coredump is actually a text file where the simulator state is
> recorded when exiting.  Note that this is always done through atexit()
> so it does not really matter *how* the simulator was exiting,
> including a caught SIGINT.
>
> We could probably even teach AVR-GDB to eventually be able to read
> these dump files in place of a normal coredump file, which allows for
> a post-mortem analysis of an aborted (e.g. through SIGINT) simulation.

Is there a typical format for this?
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. Basically that means to add 
one section to the elf which contains the memory dump. Registers are 
also included there, so I see no problems. To use elf makes it simple to 
do a postmortem analysis with symbols. In addition such a file can be 
used as a start point for continuing simulation at a frozen point, 
because reading elf is already supported. Only the ram-sections must be 
added. For reading such elf files back for continuing we have to define 
a special address range for "background registers" like 16bit timer 
temporary registers and others. Actually such registers are not visible 
via address space nor through any of the back doors.



> The avr-libc testsuite uses the option -B <addr>, aka. --breakpoint
> <addr>, in order to stop the simulation.  This option appears to be
> the same as -T now (except that -T can also handle a symbol name when
> simulating an ELF file), so maybe -B could be added as a synonym for
> backwards compatibility?  Since simulavrxx uses the same command name
> (simulavr) as the old implementation, this would be useful.

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 :-)