Re: master 695f679: Remove ; ; ; ###tramp-autoload cookie from Tramp defcustoms (Bug#47063)

[Stefan Monnier <monnier@iro.umontreal.ca>]

> Yup; if it's hoisted out of the loop, there'd be no performance impact,
> but it would make it less useful, though -- as with erroring out, you'd
> get no completion.

It'd make more sense to just keep the current (not-filtered) completion
list, so you still get completion to work.

> So I don't oppose catching errors here, really -- I'm just wondering why
> we're designing for allowing broken code in this particular
> circumstance.  Do we expect these code snippets to break for some
> reason?

I think it's because such errors have already occurred, and not being
able to `M-x` just because of a silly mistake in a "secondary" feature
is rather frustrating.
Also, this feature runs code from "all" packages, so it is inherently
exposed to a lot of potential messups.

IOW, I agree that there's no hard-and-fast reason to do that, but
I think it can make a substantial difference to the end users because
it protects them from very minor but potentially common bugs.


        Stefan