Re: New asdf version

[ Accidentally sent an unfinished version of this a few minutes ago.
Sorry.  Please disregard that one, read this one instead ]

Marco Baringer <[email protected]> writes:

> Personally i think that if you don't specify which testing framework
> to use you'll run into two problems:
>
> 1) no one will feel "required" to write tests since it's so open
>    ended.

OK, here's my reasoning.  

1) Whatever we choose will either be too heavyweight for authors of
small tools to want to get entangled with, or too underfeatured (or
just different) for authors of large systems to port their existing
test suites to.

2) asdf is a system definition tool, not a testing framework.
Incorporating someone else's testing framework into it just means
headaches tracking version skew.  If we just tell people the test
framework is available under separate cover, they have the same
installing-multiple-packages problem 

We've been talking about putting RT (from the CMU AI repository) into
SBCL contrib - it's already being used by a couple of the existing
contrib packages - if we do, I rather expect it will turn into
something of a de-facto standard at least for people using that
implementation.

3) To be absolutely honest, nobody _is_ required to write tests just
to get their package into cclan.  Sure, it'd be nice if they did, but
there's not yet sufficient social cachet to having "a package in
cclan" that we can use it as leverage to make people write anything
they weren't going to write anyway.  And in that case, I'd rather have
the untested or cursorily tested code available in cclan (and thus,
win on the ease-of-installation front) than have it randomly packaged
and stuck on a web page somewhere.

Eventually I hope that someone will produce a lint-op operation or
similar, which will assess asdf packages for good style and complain
about this kind of stuff.  But assessing taste in Lisp programming is 
probably an AI problem ...

> 2) if someone does write tests they need to also make sure the user
>    has the same testing framework (and same version) installed on the
>    local machine.

Point 2 should be a non-issue given automatic cclan-package-download
tools (RSN, ha ha).

> I really believe some standard is better than no standard and only
> ASDF (or defsystem like tools) has a chance of getting this
> accepted. [shameless plug: I have been using for a while (but have not
> yet openly distributed) a simple testing framework
> (http://www.bese.it/sw/5am.tar.bz2) which contains an implementation
> of test-op.]

I had a quick squint at this: it looks quite appealing, but I need to
read it more carefully to decide whether it is or could be a
counterexample to point (1) above.


-dan

-- 

   http://www.cliki.net/ - Link farm for free CL-on-Unix resources 


-------------------------------------------------------
This SF.net email is sponsored by:Crypto Challenge is now open! 
Get cracking and register here for some mind boggling fun and 
the chance of winning an Apple iPod:
http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0031en