Andy Cowling <[EMAIL PROTECTED]> writes:

>     lipp> I get the same error under Lucid EMACS 19.9 with BBDB 1.50
>     lipp> and VM 5.57 (beta) set to autoload.
> 
> Also happens for GNU emacs 19.22 with a 'noautoload' VM (5.57) so that
> isn't the problem.

Not quite.

The following statements were verified using bbdb-1.50 and vm-5.56
and vm-5.57, and emacs-19.22.

First, let me define some notations:

  o vm.elc-auto   is defined as vm.elc built with autoloads.
  o vm.elc-noauto is defined as vm.elc built with noautoloads.

bbdb 1.50 works with either vm.elc-auto or vm.elc-noauto!

The trick is that you have to byte-compile bbdb-vm.el
using vm.elc-noauto.

Once bbdb-vm.el is byte-compiled "properly", you can then use either
vm.elc-auto or vm.elc-noauto.

The problem seems to be that the function bbdb/vm-update-record
defined in bbdb-vm.el calls vm-start-of which is defined
as macro in vm-message.el.

If bbdb-vm.el is byte-compiled with vm.elc-auto,
then "(require 'vm)" in bbdb-vm.elc loads vm.elc-auto which
defines vm-start-of as an autoloaded macro.

It seems like to me that an autoloaded macro should be loaded
at compile time so that the macro is properly defined.

Could this be a byte-compiler problem?

Richard Y. Kim
[EMAIL PROTECTED]

Reply via email to