On Fri, 14 Apr 2000, Dave Love wrote:
> >>>>> "EZ" == Eli Zaretskii <[EMAIL PROTECTED]> writes:
>
> EZ> I hoped that it's easy to check what I asked about if you fset
> EZ> start-process to throw an error. But if that's too complicated,
> EZ> I will install your patch and try it.
>
> I'm sorry, but I've no idea what the issue is and why doing that would
> be relevant
Forget it, I just did "M-: (require 'tex-mode) RET" and didn't see any
trouble. I'm sorry I didn't do it earlier, instead of bothering you.
> I think the rules about this sort of thing need to be codified if we
> have to follow them.
I was asking a question, I didn't request to codify this, nor do I
think it's possible to codify this particular issue.
> I think it's unfortunate, though, if we can't
> generally require stuff for the compiler's (our) benefit.
If a package that provides a feature foo runs some code that calls
functions which are unbound on some platforms, then (require 'foo)
will fail on those platforms. If we want to require that feature
during compilation, I think we need either to change the require'd
package to avoid failing, or not require it at compile time and
solve whatever problems require it in another way.
