From Scott at sympoiesis.com  Sun Sep  6 21:11:16 2009
From: Scott at sympoiesis.com (Scott L. Burson)
Date: Sun, 6 Sep 2009 14:11:16 -0700
Subject: [cl-l10n-devel] Assertion failure loading CLDR
In-Reply-To: <310c21410909061018u4e766e90j95e5064f8711ef16@mail.gmail.com>
References: <310c21410909061018u4e766e90j95e5064f8711ef16@mail.gmail.com>
Message-ID: <310c21410909061411w48a7e4efjc515ceb5e10d4d1e@mail.gmail.com>

Hi,

I know very little about cl-l10n and have no plans to use it, but I am
trying to build LibCL (http://libcl.com), which incorporates cl-l10n, and
having some trouble that I thought you might want to know about.  I am
attempting the build on Allegro CL 8.1, ANSI mode (case-insensitive-upper),
with wide strings.

There is a minor problem that Allegro signals a package lock error when
cl-l10n tries to do (DEFCLASS IDENTITY ...), IDENTITY being of course a CL
builtin.  Looks like shadowing the symbol in your package would be a good
idea.

The bigger problem is the failure of assertion (SUBTYPEP CLASS 'FLEXML-NODE)
inside SAX:START-ELEMENT somewhere under (PARSE-CLDR-FILE "root").  The
problem appears to be that cldr/main/root.xml has an element (perhaps
recently added?) named "list", which becomes the LOCAL-NAME argument to
SAX:START-ELEMENT; CLASS-FOR-NODE-NAME then returns #<BUILT-IN-CLASS LIST>,
which obviously is not a subtype of FLEXML-NODE.  I would strongly suggest
that any package used for XML element names should have a (:USE) subform in
its DEFPACKAGE form so as not to inherit symbols from CL:.

I have also sent this to libcl-devel because I think they might want to
know; also, I don't understand why this doesn't fail on other platforms.

-- Scott
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.common-lisp.net/pipermail/cl-l10n-devel/attachments/20090906/e252a573/attachment.html>