[Ecls-list] deftype propagation in the ECL compiler

[Gabriel Dos Reis]

"Juan Jose Garcia-Ripoll" <juanjose.garciaripoll at googlemail.com> writes:

| On Sun, Nov 2, 2008 at 5:01 PM, Gabriel Dos Reis <gdr at cs.tamu.edu> wrote:
| > I updated just 5min ago, but it seems that I do not have all the
| > changes: I have the COERCE change, but not the CONS type change.
| > And I can confirm that a fresh ECL build failed on the testcase I
| > provided earlier.
| 
| Of course, if there is no change in src/lsp/predlib.lsp the test cases
| will fail. The change is in src/lsp/predlib, more precisely in
| register-cons-type
| (defun register-cons-type (&optional (car-type '*) (cdr-type '*))
|   (let ((car-tag (if (eq car-type '*) -1 (canonical-type car-type)))
|         (cdr-tag (if (eq cdr-type '*) -1 (canonical-type cdr-type))))
|     (if (or (zerop car-tag) (zerop cdr-tag))
|         0
|         (canonical-type 'CONS))))
| 
| What is the value of ecl/CVS/Root? Is it pointing to sourceforge?
| :pserver:anonymous at ecls.cvs.sourceforge.net:/cvsroot/ecls
| I followed the instructions at the ECL homepage and right after my
| commit the changes were available at the CVS rever :-/

Operator error :-(.  You nailed the problem down:  I have

   :pserver:anonymous at common-lisp.net:/project/ecl/cvsroot

in CVS/Root, but somehow managed to convince myself for a long time that I
was updating from the SF repository.  I can confirm that a fresh copy
from CVS repo from SF has passed the testcase, and an ECL-based
OpenAxiom is progressing past the point where it would have failed.

Sorry for the confusion.

Thanks!

-- Gaby