`eval-after-load' broken for symbols?

`eval-after-load' broken for symbols?

[Daniel Brockman]

The ubiquitous function `eval-after-load` seems to be broken.
The following code causes the message to be displayed:

    (require 'dired)
    (eval-after-load "dired"
      '(message "String works!"))

But the following code doesn’t:

    (require 'dired)
    (eval-after-load 'dired
      '(message "Symbol works!"))

From the docstring of `eval-after-load`:

    Arrange that if FILE is loaded, FORM will be run immediately afterwards.
    If FILE is already loaded, evaluate FORM right now.
        […]
    Alternatively, FILE can be a feature (i.e. a symbol), in which case FORM
    is evaluated at the end of any file that `provide's this feature.

I guess that _could_ be trying to say that "when FILE is a symbol"
it *shouldn’t* "evaluate FORM right now", but if you look at the source
code for `eval-after-load` it doesn’t actually avoid evaluating FORM,
it just happens for an unrelated reason to wrap it in a false conditional:

    (when (symbolp regexp-or-feature)
      ;; For features, the after-load-alist elements get run when `provide' is
      ;; called rather than at the end of the file.  So add an indirection to
      ;; make sure that `form' is really run "after-load" in case the provide
      ;; call happens early.
      (setq form
            `(when load-file-name
                        …

So, is this actually a bug? The behavior sure seems broken anyway.


-- 
Daniel Brockman, partner & developer
Go Interactive <http://gointeractive.se>
Twitter: http://twitter.com/dbrock
Telephone: +46706880739

Re: `eval-after-load' broken for symbols?

[Johan Bockgård]

Daniel Brockman <dbrockman@gmail.com> writes:

>     (when (symbolp regexp-or-feature)
>       ;; For features, the after-load-alist elements get run when `provide' is
>       ;; called rather than at the end of the file.  So add an indirection to
>       ;; make sure that `form' is really run "after-load" in case the provide
>       ;; call happens early.
>       (setq form
>             `(when load-file-name
>                         …
>
> So, is this actually a bug? The behavior sure seems broken anyway.

Yes. It works correctly in Emacs 23.