Re: [Gcl-devel] New random tester failures

[Camm Maguire]

Thanks again!  As you can tell, I'm trying to save on the
pre-evaluations if possible.  Simplest would be to pull all args into
a let*, but ...

I think this is fixed now, but would always appreciate your comment.
This version pre-evaluates in a let* all a) cons args whose car is not
'quote, and b) non-constant atom args which appear anywhere to the
right (at any nesting level).  This should work, right?

Take care,


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

> Camm Maguire wrote:
> > Greetings, and thanks for the quick notice!  I had introduced this
> > when speeding up certain binary operator function calls of more than
> > two arguments via compiler macros.
> > I think this should be fixed now.  Please confirm, and review if time
> > permits.  I am assuming I only need to ensure evaluation of arguments
> > which are of type cons and whose car is not 'quote.
> 
> This has apparently introduced a new bug:
> 
> Test MISC.449 failed
> Form: (FUNCALL (COMPILE NIL '(LAMBDA (A) (* 10 A (SETQ A 1000)))) 1)
> Expected value: 10000
> Actual value: 10000000.
> 
> (this test has been committed.)
> 
>       Paul
> 
> 
> _______________________________________________
> Gcl-devel mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/gcl-devel
> 
> 
> 

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