Re: [Gcl-devel] Status of test suite

[Camm Maguire]

Greetings!  Thanks for the update here.  

Personally, I feel that the work you are doing with the ansi-tests is
very useful for gcl, and should be encouraged.  We should understand
that there may be mistakes, and should attempt cross-checking with the
spec whenever possible, but the inescapable reality is that the
effect of these tests is generating movement toward making gcl more
ansi compliant -- thereby addressing perhaps its chief perceived
shortcoming in the lisp community.  It would be nice to synchronize
with clocc as time may allow, but there is such a substantial amount
of work to be done that the value of the momentum generated becomes
the most important factor.  I'd hate to see the process stalled by
discouraging remarks, bureaucratic procedures, or fine-grained
language lawyering.  There is plenty of time for that later when the
bulk of the ansi work is done, IMHO.

I like the idea of running the tests against cmucl, though, if time
permits. 

How different is your test suite from current clocc?  I'm assuming
that clocc must not be completed nor nearly so, otherwise you wouldn't
have started this project.  If that is the case, your work becomes
even more important for us, as we don't have enough people thoroughly
familiar with the spec to bring non-clocc covered issues to the fore
through their customary use, unlike the case with other lisp projects,
I imagine.

Anyway, thanks.

Take care,

"Paul F. Dietz" <address@hidden> writes:

> I posted a message about the test suite on the cmucl mailing list
> and have been getting some honest, if forceful, feedback.  Some
> of the tests are problematic.  In particular, types-4 had problems:
> it was requiring base-string to not be a supertype of string,
> (for example), but there was nothing in the standard that required
> that.  There were other problems as well.  I've fixed those I've
> been told about, but I can't be sure there aren't more.
> 
> The upshot of all this is that I've been pointedly instructed
> to work with the CLOCC test suite rather than do a bad job of
> reinventing the wheel.  I can't say I disagree with that; I
> don't want to send gcl development down a blind alley.  I may
> at some point start offering my tests to be integrated into the
> CLOCC suite rather than having a separate test suite in the gcl
> tree. I will certainly run the tests against CMUCL to weed out
> some bad ones.
> 
> What this means for you is to not trust the test suite to
> be perfect, or my understanding of the CL spec to be perfect,
> since neither are.
> 
> I'd feel worse about this if the tests I wrote hadn't already
> found two new bugs in CMUCL/SBCL.
> 
>       Paul
> 
> 
> 
> _______________________________________________
> Gcl-devel mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/gcl-devel
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah