[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [avr-gcc-list] Regression tests
Re: [avr-gcc-list] Regression tests
Tue, 15 Nov 2005 14:59:48 +0100 (MET)
> Hi Claus,
simple check out from the sources and look for the tcl scripts under
examples. There you can see how things can be activated and also
be read out. And also the gdb interface is available via normal
unix socket. So all internal and external events/states could be
set or analysed. You can write your own C/C++/Tcl/Python programs/scripts
which do the check itself. So you have the complete simulated environment
and also the full ability for all tests you need. There is no "generic"
regression test toolset, but I also have never a need for such "generic"
What I normaly do is to get a bus protocal analysis which is taken
from a 2 wire bus (clk/data) which is monitored and connected to
a lot of simulated controllers. And if the traced protocol is what I expect
nearly all things in all the controllers seems to work fine.
I use it for a model railway control system which uses lots of
avr controllers. I simulate also the behaviour of trains, the
signaling and a lot of other things. If the train arrives at
full simulated endpoint and protocoll tells that no errors occur
and the state of all control elements is in start condition after the run,
all seems to be ok. If this
is true the simulated environment has no failures and should also
work in real conditions and hardware (and it works :-).
> This sounds very exciting!
> Is there any place I can learn about it? Any examples lying around?
> On Tue, 15 Nov 2005 12:58:05 +0100, Klaus Rudolph <address@hidden>
> > Hi David,
> > I have a lot of regression tests running for my embedded code. I use
> > my simulator (simulavrxx) which also give the ability to manipulate
> > external
> > pin states and interconnections between multiple avr devices. A lot of
> > timing parameters could be measured also
> > under differnt interrupt conditions so that worst case tests could be
> > done. I think this could never work in real hardware, also jtag give
> > not full access in real time environment.
> > Bye
> > Klaus
> >> I've seen on the avr-libc-dev list that regression tests were used and
> >> on
> >> avr-freaks that one guy was building a test suite for his embedded
> >> applications but he didn't tell anything about how to do it.
> >> My question is wether it's possible or not to use regression tests for
> >> embedded applications. I can't imagine how this can work with hardware
> >> related functions. What do you use it for in avr-libc developmment?
> >> I coudn't find much information on the subject. I found some keywords
> >> like
> >> Avrora, gdb serial protocol, MUL, dejagnu but couldn't get any clear
> >> idea
> >> standing out of all this.
> >> Thanks for any clarification,
> >> David Bourgeois, still on his quest towards better programming :-)
> >> 
> >> "Build a test suite, as early as possible. Start small, and plan for
> >> growth.
> >> Add (at least) one new test for every new feature or resolved bug.
> >> Make it easy and (reasonably) quick to run, so no one has an excuse
> >> for not running it before Every check-in. I have been repeatedly
> >> rescued by the Regression Test Suite when I've inadvertantly stuck my
> >> elbow into something. Moreover, the very exercise of writing the tests
> >> can help point out holes in the specification/design."
> >> _______________________________________________
> >> AVR-GCC-list mailing list
> >> address@hidden
> >> http://lists.nongnu.org/mailman/listinfo/avr-gcc-list
> David Bourgeois
> C2ME - Centre de Conception en Microtechnologies Européen sa
> Parc Initialis - Rue René Descartes, 2 - 7000 Mons (Belgium)
> Phone : +32 65 37 42 25 - Fax +32 65 37 42 36
> AVR-GCC-list mailing list
Telefonieren Sie schon oder sparen Sie noch?
NEU: GMX Phone_Flat http://www.gmx.net/de/go/telefonie