[asdf-devel] [PROPOSAL] Easier configuration, installation, and removal of asdf systems

Krzysztof Drewniak krzysdrewniak at gmail.com
Sat Nov 20 13:06:23 UTC 2010


On Sat, 2010-11-20 at 02:10 -0500, Faré wrote:
> On 19 November 2010 16:20, Krzysztof Drewniak <krzysdrewniak at gmail.com> wrote:
> >> 2- It's not clear that anyone would use it. What problem are you solving?
> >>  How does it help?
> >>
> > If solves the problem of not having a standard way to "install" a common
> > lisp library. See my other reply for details.
> >
> There is nothing whatsoever to install in a CL library.
> FASLs? They vary from implementation to implementation,
> and even in a given implementation, depending on
> OPTIMIZE settings, *features* and possibly more
> (code coverage, grovelling, some lisp-in-lisp compilation
> for e.g. continuations, etc.)
> 
> And what the hell do you configuration settings correspond to?
> Yet another layer of bureaucracy between the user and
> the above-mentionned settings and features?
> 
> > As fetching dependencies is clearly not in scope for asdf, dependencies
> > will be checked to see if they are loadable in the beginning of
> > install-op, and if they are not present, a helpful error is signalled.
> >
> What when a dependency is compiled with a different settings, entailing
> different semantics for the final "library"?
> 
> >> 4- How do you distinguish between fasls compiled with different options.
> >>
> > If config.asdf is modified, fasls are recompiled. If two different
> > source trees of the same system are installed, they will vertanly not be
> > installed in the same place, thuse distinguishing the fasls.
> >
> That's not the question asked. How do you make sure there is no unwanted
> sharing between two "versions" of the fasls for a same library as installed
> in the same tree?
> 
> > Also, if you are using different "trees" of libraries, either use
> > different :registry files, or don't use these operations.
> >
> That's not the question asked. If you have different trees,
> you don't need to distinguish configurations.
> But usually you don't.
> 
> Or what the hell are your "configurations" useful for?
> 
> > To clarify something in advance, config-get does NOT require you to have
> > used install-op, you could have done the installation manually.
> >
> Your whole proposal sounds like more bureaucracy that doesn't actually
> help anyone get anything done but adds more work for everyone.
> I say emphatically NO to any of it.
> 
> Please provide a use case — we'll tell you how to do without it.
> 
> There are ways that a build system could benefit from understanding
> multiple "configurations" of a same tree, but ASDF will never ever
> possibly do that, since that would require major refactoring and
> breaking backward compatibility. If you're interested in such things,
> consider hacking XCVB instead.
> 
> [ François-René ÐVB Rideau | Reflection&Cybernethics | http://fare.tunes.org ]
> Big Business has to *sell* the Vietnam War to you. Big Government can just
> draft your ass and ship you overseas under threat of imprisonment and/or
> death. Now, which one is better again? — rho at netdoor.com

Never mind. asdf-install already solved most of the problem.
-- 
X-Real-Email-With-Antispam: krzysdrewniak at gmail dot com
pgp key on keyserver.ubuntu.com 94F02AE8

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part
URL: <https://mailman.common-lisp.net/pipermail/asdf-devel/attachments/20101120/8441f03b/attachment.sig>


More information about the asdf-devel mailing list