Re: Fixing numerous `message' bugs..

[Dave Goel]

>
>> Here's an idea: make up an argument to `message' that says treat the
>> next argument literally.  How about t?
>>
>>    (message t STRING) is equivalent to (message "%s" STRING)
>>    (message t nil) is equivalent to (message nil)
>>
>> So in the cases that are buggy it suffices to add t as the first argument.

The above is the neatest idea so far.  It does not involve a new
function, and makes fixing code really easy.  And, any author that desires
a literal treatment can simply use (message t string).

> I don't particularly like it.  What about (message t nil 7), what is
> that supposed to return?  nil too? 

Sure, why not?  ((Or, it could return an error. ))

> Anyway, there is little point to make this specific to message.  So
> if at all, we would special-case format instead.  Which means that
> (stringp (format ...)) is no longer guaranteed to be true.


Can you elaborate?  Why should a `message' bug cause changes to
something as general as format?  And, how will that fix the problem at hand?

Besides, tweaking `format' will further our non-compatibility with
Common Lisp.