Re: loading functions from a shared library

Re: loading functions from a shared library

[William M. Perry]

[ Thread about using sysdll.c and sysdll.h from XEmacs for loading shared
  libraries in Emacs .... ]
Dave Love <d.love@dl.ac.uk> writes:

>>>>>> Richard Stallman writes:
>
>  > It could be that the people who wrote those file signed FSF legal
>  > papers for this code.
>
> For what it's worth, the basic code I got a couple of years ago was OK
> (ex Bill Perry) but I don't know what's happened to it since.  That code
> would be easy enough to write from scratch anyway, but is basically doing
> the job libtool abstracts.  (I dare say libtool didn't at the time Bill
> started.)

Sorry for the delay in responding - this got buried in the 'to be answered'
folder.  *sigh*

Correct, libtool did not support this type of abstraction at the time.  :)
Nowadays, I would say we should just use the 'ltdl' library that libtool
provides.  It does this and a fair bit more.

>  > The only way to find out if to talk with people who maintained the
>  > code and try to form a complete list of contributers to these files.
>
> I've sometimes been able to get convincing info out of the XEmacs CVS
> logs in the past, but often not.

The diffs are a better source for this type of thing in some cases.  Diffs
from the 1.1 version in CVS show the only changes were adding the
boilerplate, and some formatting changes along the lines of:

-dll_func dll_function(dll_handle h,CONST char *n) {
-  return(NULL);
+dll_func
+dll_function (dll_handle h, const char *n)
+{
+  return NULL;

But with ltdl being available and easily distributable with Emacs if
someone doesn't have it installed, I would say it is a moot point.

-bp
-- 
Ceterum censeo vi esse delendam

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://mail.gnu.org/mailman/listinfo/emacs-devel

Re: loading functions from a shared library

[Dave Love]

wmperry@gnu.org (William M. Perry) writes:

> Correct, libtool did not support this type of abstraction at the time.  :)
> Nowadays, I would say we should just use the 'ltdl' library that libtool
> provides.  It does this and a fair bit more.

Yes, though I found it non-trivial to figure out how to drive libtool
for this purpose, and I don't understand some behaviour of the result.

> > I've sometimes been able to get convincing info out of the XEmacs CVS
> > logs in the past, but often not.
> 
> The diffs are a better source for this type of thing in some cases.

I found they're necessary, actually, but still not necessarily good
enough in general :-(.  (I laboured on this sort of thing quite a lot,
and I doubt there's a lot more of interest to look at.)

> But with ltdl being available and easily distributable with Emacs if
> someone doesn't have it installed, I would say it is a moot point.

I don't think the API is necessary either, at least for loading
modules per se.

The implementation I have just extends `load'.  It needs work on
understanding a couple of issues, snarfing doc strings and packaging.

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://mail.gnu.org/mailman/listinfo/emacs-devel