Re: [PATCH] Avoid `SCM_VALIDATE_LIST ()'

[Ken Raeburn]

On Sep 7, 2008, at 11:24, Neil Jerram wrote:
> > I'm using a Mac as my main machine these days.  The Emacs unexec  
> > mechanism
> > interacts badly with the Mac version of malloc, so Emacs uses its own
> > version specially tailored to use Mac system allocation routines in  
> > a way
> > that works with unexec.  Guile doesn't have any hooks for doing  
> > that sort of
> > thing.  Of course you can configure with -Dmalloc=unexec_malloc and  
> > so on,
> > but then you're guaranteed not to be able to build an actual guile
> > interpreter executable without the extra Emacs code.
>
> Not sure I understand.  Surely you would always know if you're
> building with or without the Emacs code, and so could define malloc or
> not accordingly?

Not if you want to build and install the full guile package and then  
link emacs against libguile, which is how I've generally done it.  Or  
if you want to build both emacs and guile executables at once so you  
can test libguile without the rest of emacs.  Of course, I could just  
drop a copy of guile into the source tree and tweak the build process  
to skip the 'guile' binary, or add wrapper functions that only link  
into that executable and cause the redirected calls to invoke malloc  
and friends after all.

I could also create function-pointer variables that could be changed  
before guile initialization, and I may do it that way for my own use,  
but it's not clear to me that there'd be any value in feeding that  
back upstream.

Ken