Re: Guile REPL Server

[Jonas Hahnfeld]

On Thu, 2022-05-19 at 13:28 +0200, David Kastrup wrote:
> Jonas Hahnfeld via LilyPond user discussion <lilypond-user@gnu.org>
> writes:
> > On Tue, 2022-05-17 at 23:28 +0200, David Kastrup wrote:
> > > Personally, I'd rather have it.  The tiny fraction that will want
> > > readline support will want it for running scheme-sandbox or other forms
> > > of REPL.  It greatly increases the friendliness of playing around with
> > > Scheme stuff in the manner the manual suggests, and the people playing
> > > around with Scheme stuff are the most likely candidates for becoming
> > > future contributors.
> > 
> > As I just replied in my other message, the REPL itself works fine.
> 
> "[readline] greatly increases the friendliness of playing around with
> Scheme stuff in the manner the manual suggests, ..."

Ah okay, in my (non-native) understanding "it" referred to the REPL in
general which is more friendly than putting statements in a file and
compiling it...

> > This paragraph is outdated and I'm working towards not shipping the
> > separate guile executable with the official binaries.
> 
> Probably by recoding lilypond-invoke-editor in a different language?

Yes, it's the only script not written in Scheme whereas all others are
Python, see https://gitlab.com/lilypond/lilypond/-/issues/6311

> > I'll update the documentation afterwards to mention that it can be
> > installed on many Linux systems, but is / will not be included in the
> > binaries.
> 
> Does this have to do with byte-compilation?

Partly: to use the guile executable, you'd need to set at least
GUILE_LOAD_PATH and GUILE_LOAD_COMPILED_PATH (which we currently do in
the wrapper script for lilypond-invoke-editor). Other than that, I
don't think it's a good idea to distribute unused stuff, in particular
since static linking means the guile executable will duplicate the
entire libguile that is already linked into lilypond.

Jonas

signature.asc
Description: This is a digitally signed message part