Re: [avr-libc-dev] Automated testing project

[Peeter Vois]

As my proposed testsuite uses gdb serial protocol maybe it would be
possible to set up hardware testplatforms for all / missing devices?
This would need only minimal changes into testsuite.

We could improve the simulavr too in this sence if we will have more
resources.

Having regression test running on some instances of devices from device
classes would be step forward too.

I would like to make the following steps in the near future ( in the
order of importance ): 
1. Import the MUL modifications into CVS
2. go on with the log() optimisations
3. improve the testsuite to have statistical module for results
4. Import the testsuite into CVS somewhere
5. I remember that I had some year ago porblem with constant + (nan) I
would check this too

Peeter



Ühel ilusal kenal päeval Tue, 18 Oct 2005 23:36:39 +0200
Joerg Wunsch <address@hidden> kirjutas:

> As Peeter Vois wrote:
> 
> > I'm thinking what would I do with the testing environment. We could
> > merge it with avr-libc project or create a new project for libc
> > automated testing.
> 
> I'm quite busy right now, and wish I'd already made avr-libc-1.4.0
> reality.
> 
> Anyway, after that happened, it already became pretty clear to me
> we need some sort of regression test.  Currently, for any new
> device, I'm running a small test program on it to see whether the
> library adaptation for the new device actually works.
> 
> After seeing how GCC uses dejagnu, I've already thought about using
> that for regression tests, but I'm open for any option here.
> 
> One major showstopper could be the limited device support in
> simulavr/simulavrxx though.  Basically, when extending the library
> for new devices, it would require the simulation to also cover that
> new device in order to make this a useful option.
> 
> -- 
> cheers, J"org               .-.-.   --... ...--   -.. .  DL8DTL
> 
> http://www.sax.de/~joerg/                        NIC: JW11-RIPE
> Never trust an operating system you don't have sources for. ;-)
>