bug#37517: New rx.el breaks a number of packages - missing symbol rx-submatch-n

[Eli Zaretskii]

> From: Mattias Engdegård <mattiase@acm.org>
> Date: Thu, 26 Sep 2019 12:01:46 +0200
> Cc: Clément Pit-Claudel <cpitclaudel@gmail.com>,
>  Tim Cross <theophilusx@gmail.com>, fmdkdd <fmdkdd@gmail.com>
> 
> Thanks for the report. The old rx code was apparently written prior to the 
> double-hyphen convention for marking package-internal symbols, and 
> `rx-submatch-n' is clearly an internal function; I'm not sure why it is used 
> by flycheck.
> 
> Of course we could attempt to implement `rx-submatch-n' and what other 
> internal stuff happen to be used by external packages, but that's an 
> open-ended endeavour. Furthermore, it may not always be possible to replicate 
> the expected semantics (internal variables are particularly difficult in that 
> respect).
> 
> I'd much rather work with package maintainers to promote the use of stable 
> interfaces. What about the attached patch (untested)?

At the very least, this information, including how to emulate
functions you removed, should be in NEWS.  Bonus points for providing
the missing APIs, even if you consider them internal.  The package is
so old that by now we should consider its APIs to be a de-facto
standard.  A re-implementation that claims to be compatible cannot
just toss them.

Thanks.