[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: More over-engineering
From: |
David Kastrup |
Subject: |
Re: More over-engineering |
Date: |
Fri, 27 Nov 2015 18:37:20 +0100 |
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/25.0.50 (gnu/linux) |
Aurélien Aptel <address@hidden> writes:
> On Fri, Nov 27, 2015 at 6:00 PM, Stefan Monnier
> <address@hidden> wrote:
>> Are we really allocating a structure for every Lisp_Object value we pass
>> through the modules API? Why do that?
>
> Look at the old thread for Philip's arguments. As far as I'm
> concerned, we need an indirect pointer to support wide-int
> Lisp_Objects on 32bit platforms. Global reference can still be made
> with the current code. We have a reference counting hash-table to keep
> track of them. Non-global refs are marked explicitely in case the GC
> misses them.
If the same modules cannot be used at the same time for wide-int and
normal int Emacs compilations (can they?), then it would seem that
module code could become significantly simpler when declaring wide-int
and module support mutually exclusive.
--
David Kastrup
- More over-engineering, Stefan Monnier, 2015/11/27
- Re: More over-engineering, Aurélien Aptel, 2015/11/27
- Re: More over-engineering,
David Kastrup <=
- Re: More over-engineering, Stefan Monnier, 2015/11/27
- Re: More over-engineering, Eli Zaretskii, 2015/11/27
- Re: More over-engineering, Stefan Monnier, 2015/11/27
- Re: More over-engineering, Eli Zaretskii, 2015/11/27
- Re: More over-engineering, Stefan Monnier, 2015/11/27
- Re: More over-engineering, Eli Zaretskii, 2015/11/28
- Re: More over-engineering, Stefan Monnier, 2015/11/28
- Re: More over-engineering, Eli Zaretskii, 2015/11/28
- Re: More over-engineering, Paul Eggert, 2015/11/28
- Re: More over-engineering, Stefan Monnier, 2015/11/30