Re: PATCH: use a slave interpreter

[Rob Savoye]

On Mon, Feb 09, 2004 at 06:50:17AM -0500, Michael Elizabeth Chastain wrote:
 
> If the slowdown is approximately 4% then that will be okay with me.  In
> the long run, Moore's Law gives us 4% per month, and it's okay with me
> to allocate 2% per month for toolchain expansion.  Most of that goes to
> gcc.  :(  If dejagnu does something this expensive once per year then it
> will be okay.

  Testing is not a fast process if you want it to be accurate. The ability
to have slave interpreters is potentially useful, as I assume we could also
invoke a python, or java, or whatever interpreter if we wanted...
 
> with this feature, and other people are still using sourceware
> dejagnu which does not have this feature.  It would be really
> helpful to retire the sourceware version of expect+dejagnu before
> deploying anything like this.

  We don't have any access to the sourceware version. None of us actively
maintaining DejaGnu work for Cygnus/Redhat anymore... I do know folks there
seem to prefer having a version that changes less often to simplify their
release process... but maybe it should be internal to Redhat and GCC/GDB
(etc...) developers, but not so accessible to everyone else. Ideally Redhat
should just adopt working out of the main CVS repository on subversions.

        - rob -