Re: Customization problem

[Richard Stallman]

    If we decide to protect the customizations from being cleared
    accidentally, we must prevent the situation that a file which contains
    the application of `custom-set-variables' is to be loaded but the
    application of `custom-set-variables' in the file is not evaluated
    (because of errors, for example) happening.

That is mostly right.  We can't to prevent that from happening,
but it is enough to record that it happened, so we can disallow
saving if it happened.

(I think that's what you actually mean.)

    [1] The application of `custom-set-variables' is evaluated before any
        other forms in a file.

It would be both difficult and incorrect to evaluate them out of order.

    [2] If a file contains the application of `custom-set-variables', the
        file is identified.  If the application of `custom-set-variables'
        in the file is not loaded, let `custom-save-variables' signal
        errors.  Otherwise, `custom-save-variables' just behaves as usual.

That kind of approach is what I have in mind.

    [3] If a file contains the application of `custom-set-variables', any
        error in the evaluation of the forms before the application is
        ignored so that the application of `custom-set-variables' is
        guaranteed to be evaluated.

That would be incorrect and give strange results.

    [4] The mechanism of customizations saving is changed so that the
        application of `custom-set-variables' is guaranteed to be
        evaluated.

I can't see how that could possibly be coherent, though.

    For [1] and [2], searching every file for the application of
    `custom-set-variables' is a waste.

What do you mean by "every file"?  We know which file is supposed
to have the custom-set-variables call.  It is the one speciifed
by custom-file.

Do you want to work on [2]?