[Openmcl-devel] 0.14 darwin build procedure?

Gary Byers gb at clozure.com
Fri Jul 4 21:43:48 PDT 2003


There was a bug(let) introduced in NTHCDR a couple of days ago, which
caused it (and NTH) to try to handle the general case of an UNSIGNED-BYTE
index (instead of simply assuming that any valid list index must be a fixnum)
when the index was 0.  The unsigned-fixnum and the UNSIGNED-BYTE case should
be entirely equivalent, but very early in the loading sequence the
predicate UNSIGNED-BYTE-P is undefined; NTHCDR should have realized that
0 is a fixnum and never reached this point.

The code that triggered the buglet for you was trying to update a method
cache; those caches are sort of simple hash tables whose layout depends
at least partly on the random-number generator (and may depend on other
not-entirely deterministic factors.)  I think that this means that you've
been running into this more because of bad luck than for any other reason.

On Fri, 4 Jul 2003, Jon Buffington wrote:

> I disabled my init file and rebuilt. (BTW, Is there a switch in
> ./dppccl that disables init file loading? If not, can I put a request
> -noinit?)

-n or --no-init will suppress loading the init file; -h or --help will
show this and other command-line arguments.

> Below is the output from the debugger that you requested. I
> will also re-download the binaries and source to re-build the image
> before doing a cvs update.

In this particular case, that'd have told us that it was a recent CVS
change that was causing the problem.  (I stared at NTHCDR for 10 minutes
before seeing the problem and another 10 minutes before seeing that the
bignum case didn't work and was grossly inefficient as well, so it's
not clear that that would have helped as much as the backtrace/register
info did.)

>  I was trying to make sure I included your
> recent fixes prompted by Kevin's tests. Is there a particular cvs tag I
> should use to receive blessed fixes? For example, in PLT scheme the
> 'exp' tag is guaranteed to build the latest source.

Not having run into the NTHCDR problem or having heard of it, I would
have said that the latest and greatest stuff on the Pre_0_14 tag would
have built cleanly (and that this should generally be true most of the
time); if there were some scheme to have blessed the CVS sources as of
an hour ago, I would have blessed them.

There clearly has to be -some- scheme to keep the "so-bleeding-edge-
it's-not-even-parenthesized-right" (or, more likely, "very-hard-to-
bootstrap-from-latest-binaries") stuff separate from the "latest-
believed-to-be-mostly-working-please-test-it" stuff.  At the moment,
the Pre_0_14 CVS head is supposed to be in the in the latter category,
and it seems that you just wound up unexpectedly testing NTHCDR.

>
> Cheers,
> Jon

_______________________________________________
Openmcl-devel mailing list
Openmcl-devel at clozure.com
http://clozure.com/cgi-bin/mailman/listinfo/openmcl-devel



More information about the Openmcl-devel mailing list