On 5 April 2017 at 04:28, Ben Pfaff <[email protected]> wrote:

> On Tue, Apr 04, 2017 at 01:07:58AM +0100, Reuben Thomas wrote:
> > [Ben: Ccing you on Bruno Haible's advice]
> >
> > Having gone through the changes I needed to make to my sources to use the
> > relocatable-lib-lgpl module, here's a list of things I think should be
> > documented in doc/relocatable-maint.texi. If they're agreed to be
> correct,
> > I'll prepare a documentation patch.
>
> I think that Bruno is really the expert here.  I helped out with
> "relocatable" to a significant degree, but I don't have any expertise
> with libraries.  If Bruno approves, then I second it.
>

​Thanks, Ben.

Bruno (& Ben if you're still listening!), I would like to suggest a
simplification:

1. We should drop the --enable-relocatable flag, unless there's some
problem which is unrelated to performance (which at least without
ENABLE_COSTLY_RELOCATABLE takes a negligible hit on all platforms).

2. We should go further, and remove the ENABLE_COSTLY_RELOCATABLE define.
As far as I can see, this only applies to Cygwin, and these days there are
plenty of alternatives for building free software on Windows. If you want
maximum fidelity to a POSIX system, and can't or won't use Microsoft's
latest "built-in GNU/Linux", then fine, use Cygwin, but since it translates
a lot of mechanisms on the fly between Windows and POSIX, you can expect
some slowdown. In any case, it's a one-off for us, on program start-up.

If you're happy with that suggestion, my documentation task will be
simpler, and I'll also of course provide suitable patches to remove the
configure flag and define, plus their documentation.

For backwards compatibility, ENABLE_RELOCATABLE would still be defined.

-- 
http://rrt.sc3d.org

Reply via email to