Re: [Sbcl-devel] Add a make-load-form for the sb-thread:mutex type?

[Lars R. N. <lar...@gm...>]

On Mon, 2009-05-18 at 16:57 +0300, Nikodemus Siivola wrote:
> 2009/5/18 Lars Rune Nøstdal <lar...@gm...>:
> 
> > Maybe a MAKE-LOAD-FORM for the SB-THREAD:MUTEX type should be added? I
> > must admit that I do not fully understand how this works or whether it
> > could create problems, but anyway:
> 
> Not a good idea, I think. Consider:

I agree. I totally expected DEFCONSTANT to only consider the top-level
reference to the CONS to be constant -- so forget I ever mentioned
this.


> 
> >        (defconstant +id-counter+
> >          (cons 0 (sb-thread:make-mutex)))
> 
> This is a bad idea: SBCL will evaluate the initialization form both a
> compile and load-time, as explicitly allowed by DEFCONSTANT: this
> means you will get an error at load time, and if you accept the new
> value you will end up with a new lock, which is probably not what you
> want.

Yes, I was only looking at the behavior of COMPILE-FILE here. I.e., I
did not call LOAD.


> You want either DEFVAR or DEFGLOBAL.

Yeah. SB-EXT:DEFGLOBAL it is. :)

CL:DEFCONSTANT is much "stronger" than I thought. I was expecting
CL:DEFCONSTANT to behave somewhere in-between CL:DEFCONSTANT and
SB-EXT:DEFGLOBAL.

For reference, here is the link you mentioned on IRC:
http://www.lispworks.com/reference/HyperSpec/Body/03_bd.htm


> 
> Cheers,
> 
>  -- Nikodemus
>