Re: [PATCH] [WIP] Port feature/native-comp to Windows.

[Eli Zaretskii <eliz@gnu.org>]

> From: Nicolas Bértolo <nicolasbertolo@gmail.com>
> Date: Sat, 9 May 2020 13:07:44 -0300
> Cc: Andrea Corallo <akrl@sdf.org>, emacs-devel@gnu.org
> 
> > Can you tell the details?  Where does the freloc table come into play
> > in this context?
> 
> It does not. It was a question to Andrea because I do not understand why we
> can't use the OS dynamic linker. It is not related to the MSYS issue.
> The freloc table stores a bunch of function pointers to internal Emacs
> functions.
> I think it would be cleaner to just declare each of those functions as an
> "IMPORTED" function, which is equivalent to adding a line like this to
> the generated code:
> 
> extern T a_function(T1, T2, ... Tn);
> 
> Then the dynamic linker will wire things together when loading the eln file.
> Maybe I am missing something about how dynamic linking works.

There's an issue here due to the differences in how shared libraries
are loaded on Posix platforms and on MS-Windows.  Andrea probably
doesn't know how this is done on MS-Windows.

I'm sorry to ask questions about the native-comp branch, I don't
remember the details: how does it load the *.eln files?  Does it use
dlopen and dlsym?  If so, we should do the same on Windows, which
means functions from *.eln will be called via a function pointer and
not directly.

Calling functions from *.eln files directly is less desirable, as that
will require import libraries and other similar atrocities.