> Thanks for making an updated patch. Got around to trying it out. The
> patch failed at one point, but the stuff before and after (with manual
> tweaking) worked ok. I think I've correctly applied the failed part.
yes. It's just changing a +86,7 into a +85,7 at the line patch doesn't
like -- I removed an added empty line and didn't change the numbers too.
> Anyway, it seems to work fairly well, but fails a fair number of the
> tests. Most of the tests that involve floating-point results seem to
> have single-precision results. (I think this is a bug in openmcl
> because (integer-decode-float 1d20) returns results that match what
> CMUCL/Clisp say for (integer-decode-float (float 1e20 1d20)).)
That was a bug in openmcl's reader, thanks for mentioning that. I sent a
bugfix for it to the openmcl people.
> There are some other failures that I can't remember right now. Does
> this match the kinds of results you get?
Yes. There are still failures in rtest3(70), rtest8(24), and rtest12.
rtest12 is the issue with the compiled rules, don't know about the others.
> Since gcl/cmucl/sbcl/clisp all seem ok, I can check this patch in if
> you like.
That's fine with me.
Regards,
Yannick