bug#72313: 31.0.50; Warning about cl-member possibly being undefined when using cl-pushnew

[Tassilo Horn <tsdh@gnu.org>]

Arash Esbati <arash@gnu.org> writes:

> Stefan Monnier via "Bug reports for GNU Emacs, the Swiss army knife of
> text editors" <bug-gnu-emacs@gnu.org> writes:
>
>>> So the conclusion is to simply (require 'cl-lib) instead of
>>> (eval-when-compile (require 'cl-lib))?  But how can the
>>> byte-compiler conclude that cl-member might not be defined at
>>> runtime given that it's an autoloaded function?
>>
>> It's autoloaded from `cl-seq` when `cl-lib` is loaded.
>> It's not autoloaded at startup.

What's that magic?

  emacs -Q
  (featurep 'cl-lib) ;=> nil
  (describe-function 'cl-member) ; prints the help string to *Messages*
  (featurep 'cl-lib) ;=> t

Do the help facilities load stuff on demand?

> How about this approach then?
>
> --8<---------------cut here---------------start------------->8---
> diff --git a/tex-info.el b/tex-info.el
> index dc69762e..0f77e057 100644
> --- a/tex-info.el
> +++ b/tex-info.el
> @@ -25,7 +25,8 @@
>  ;;; Code:
>
>  (eval-when-compile
> -  (require 'cl-lib))
> +  (require 'cl-lib)
> +  (declare-function cl-member "cl-seq"))
>
>  (require 'tex)
> --8<---------------cut here---------------end--------------->8---

Let's just require cl-lib.  declare-function only silences the
byte-compiler and the usage of cl-member at runtime is a fact; no
cheating allowed.  It's also an implementation detail we should not
bother with.  Tomorrow, Stefan might have optimized it to use some
cl-memql-1-wazzup for the specific case we have here. :-)

Bye,
  Tassilo