Re: [PATCH] Require full GMP for big integer module functions.

[Paul Eggert]

Philipp Stephani wrote:

> if emacs-mini-gmp.h ever
> becomes incompatible, the behavior will silently become undefined,

There is no emacs-mini-gmp.h so I assume you mean Emacs's src/mini-gmp.h.  That 
file is identical to GMP's mini-gmp.h, so isn't the Emacs-relevant 
incompatibility the same as the incompatiblity that the GMP maintainers are 
worried about?
> My understanding is that emacs-mini-gmp.h is
> maintained as part of Emacs

They're identical to the GMP mini-gmp.c and mini-gmp.h so they're primarily 
maintained as part of GMP. The only part relevant to Emacs maintenance is that 
when we update Emacs src/mini-gmp.h and src/mini-gmp.c from GMP sources, we 
should look for Emacs-relevant incompatibilities and if there are any (which 
typically there are not) we should delay these updates until the next Emacs 
release. But this is something we should do anyway - it's an ordinary part of 
release maintenance.

There's an analogy to Gnulib here. At some point after forking emacs-26 from the 
master branch, we stopped propagating Gnulib patches into emacs-26 because they 
were considered too likely to break what was intended to be a stable branch. If 
mini-gmp had been part of Emacs 26, I expect we would have taken the same 
approach to mini-gmp as well.

Which reminds me, now's as good time as any to sync Emacs master to the latest 
GMP copies of mini-gmp.h and mini-gmp.c, so I did that.