cclan-list — Discussions of all things CCLAN

Re: [cclan-list] aside from the pangs of history, what about moving ASDF to common-lisp.net

[Nikodemus S. <nik...@ra...>]

On Wed, Jan 14, 2009 at 5:31 PM, Robert Goldman <rpg...@si...> wrote:

> One of the issues from my correspondence with Gary leading up to this
> was that the people who are administrators for the Sourceforge project
> may no longer wish to be performing that task.

If this is the case, then it is best stated out loud in public. (Or at
least verified from the individuals involved, instead of assuming so.)

> Reading between the
> lines, if Gary is going to take more of that role, he might be more
> comfortable doing it on c-l.net.

Ditto.

Apropos, I was moderately surprised to discover that Richard Kreuter
doesn't hold the CCLAN commit bit currently.

Cheers,

 -- Nikodemus

Re: [cclan-list] aside from the pangs of history, what about moving ASDF to common-lisp.net

[Robert G. <rpg...@si...>]

Nikodemus Siivola wrote:
> On Tue, Jan 13, 2009 at 2:49 AM, Gary King <gw...@me...> wrote:
>> Are there good reasons to keep ASDF on sourceforge?
>>
>> If not, then what do folks think about moving it to common-lisp.net?
> 
> Getting this back onto topic...
> 
> Since you're the one asking, can you provide your own Pro/Con analysis
> -- even if your Con half says just "None that I can see."? :)
> 

One of the issues from my correspondence with Gary leading up to this
was that the people who are administrators for the Sourceforge project
may no longer wish to be performing that task.  Reading between the
lines, if Gary is going to take more of that role, he might be more
comfortable doing it on c-l.net.

best,
r

Re: [cclan-list] aside from the pangs of history, what about moving ASDF to common-lisp.net

[Nikodemus S. <nik...@ra...>]

On Tue, Jan 13, 2009 at 2:49 AM, Gary King <gw...@me...> wrote:
> Are there good reasons to keep ASDF on sourceforge?
>
> If not, then what do folks think about moving it to common-lisp.net?

Getting this back onto topic...

Since you're the one asking, can you provide your own Pro/Con analysis
-- even if your Con half says just "None that I can see."? :)

Cheers,

 -- Nikodemus

Re: [cclan-list] aside from the pangs of history, what about moving ASDF to common-lisp.net

[Gary K. <gw...@me...>]

Hi Robert,

As someone who never used subversion, I stand corrected (actually, I'm  
sitting). Now that you mention these features, I do remember that they  
are big win over CVS. I think the half-formed thought I was recalling  
was that now that good dvcs systems are here, moving from CVS to  
subversion seems like too small a step. I.e., that if one is going to  
move away from CVS, then it would be better to move to some DVCS.

> I feel crotchety in the same way GP lisper does, not so much because I
> dislike DVCS per se, but because of the babel of DVCS's.  bazaar, git,
> arch, darcs, what a pain.  It annoys me when the ratio of the number  
> of
> different DVCS's I need to know to the number of packages I actively
> modify gets too close to one.

Oh yes. I strongly agree. Sometimes it's "I want to try X but I'll be  
damned if I'm going to install another DVCS."

> On a project by project basis, I think it would be nice to achieve the
> following level of rationality:  (1) admit that the use of DVCS's come
> at a higher cost in terms of complexity (how significant the cost is
> depends on the community --- e.g., to linux kernel hackers, the cost
> would be essentially zero because it's already been paid); (2)  
> consider
> what benefit a DVCS offers in exchange for the higher complexity cost
> and (3) roughly speaking, subtract (1) from (2), and use that to  
> inform
> the decision.

Yes.

Re: [cclan-list] aside from the pangs of history, what about moving ASDF to common-lisp.net

[Gary K. <gw...@me...>]

> Please leave a method for those folkes interested in a silly idea,
> they just want to utilize the software, not master a rcvs system to
> get it.  It used to be that a hyperlink to a tar file download was
> pretty common, but recently those have disappeared.
> Just include one.


<possibly-sad-humor>
Hey, if you're not willing to master a host of arcana just to do  
program in Lisp, then why are you here
</possibly-sad-humor>

I completely agree. It's crazy to force folks to get the RCS du jour  
just so that they can try system Z.


--
Gary Warren King, metabang.com
Cell: (413) 559 8738
Fax: (206) 338-4052
gwkkwg on Skype * garethsan on AIM * gwking on twitter

Re: [cclan-list] aside from the pangs of history, what about moving ASDF to common-lisp.net

[GP l. <fp...@cl...>]

   From: Gary King <gw...@me...>
   Date: Tue, 13 Jan 2009 08:39:55 -0500

   > Actually, I'm rather tired of the current "but it's better" nonsense
   > on version control systems.  I've heard a dozen names now, and most of
   > that crap is hard to maintain and disappears quick to the next-new-kid
   > on the block.  ...
   > Some new tool that will be dead
   > next week makes sure that software dies.


   I agree with your conclusions but not your premises <smile>. Many of  

Boy am I surprised ;-)

   the dvcs tools have been around for a long time (in software years),  
   are used by zillions of people and work much better than CVS for lots  
   and lots of use cases. OTOH, I no longer see the point of Subversion  
   since it, at least, is not better enough than CVS.

   If you do decide to move and do decide to use a DVCS, then I'd vote  

ah, I thought YOU were asking, I'm just a run-over bystander

   that we use git (I'm personally a big darcs fan but bit has much  
   greater momentum, is "more" cross-platform and is building a bigger  
   echo-system, ummm, ecosystem.) Other thoughts?

Just one

Please leave a method for those folkes interested in a silly idea,
they just want to utilize the software, not master a rcvs system to
get it.  It used to be that a hyperlink to a tar file download was
pretty common, but recently those have disappeared.
Just include one.

r

Re: [cclan-list] aside from the pangs of history, what about moving ASDF to common-lisp.net

[Robert G. <rpg...@si...>]

Gary King wrote:
> Hey,
> 
>> Actually, I'm rather tired of the current "but it's better" nonsense
>> on version control systems.  I've heard a dozen names now, and most of
>> that crap is hard to maintain and disappears quick to the next-new-kid
>> on the block.  ...
>> Some new tool that will be dead
>> next week makes sure that software dies.
> 
> 
> I agree with your conclusions but not your premises <smile>. Many of  
> the dvcs tools have been around for a long time (in software years),  
> are used by zillions of people and work much better than CVS for lots  
> and lots of use cases. OTOH, I no longer see the point of Subversion  
> since it, at least, is not better enough than CVS.

I'd have to disagree with you here.  I have come to love SVN.  The
ability to move files, the notion that commits are coherent changesets,
rather than onesy-twosy mods to single files that are difficult to
coordinate ex post facto without the aid of tags, and the improvements
to tagging and branching put it head and shoulders above CVS.  And since
the interface is almost identical to CVS's, the changeover involves
little or no training.
> 
> If you do decide to move and do decide to use a DVCS, then I'd vote  
> that we use git (I'm personally a big darcs fan but bit has much  
> greater momentum, is "more" cross-platform and is building a bigger  
> echo-system, ummm, ecosystem.) Other thoughts?

I feel crotchety in the same way GP lisper does, not so much because I
dislike DVCS per se, but because of the babel of DVCS's.  bazaar, git,
arch, darcs, what a pain.  It annoys me when the ratio of the number of
different DVCS's I need to know to the number of packages I actively
modify gets too close to one.  Especially when these things are
significantly more complicated to use than a standard VCS.  If git drove
the rest off the map (or way out in the periphery), then the changeover
would be a lot more appealing....

On a project by project basis, I think it would be nice to achieve the
following level of rationality:  (1) admit that the use of DVCS's come
at a higher cost in terms of complexity (how significant the cost is
depends on the community --- e.g., to linux kernel hackers, the cost
would be essentially zero because it's already been paid); (2) consider
what benefit a DVCS offers in exchange for the higher complexity cost
and (3) roughly speaking, subtract (1) from (2), and use that to inform
the decision.

cheers,
r

Re: [cclan-list] aside from the pangs of history, what about moving ASDF to common-lisp.net

[Gary K. <gw...@me...>]

Hey,

> Actually, I'm rather tired of the current "but it's better" nonsense
> on version control systems.  I've heard a dozen names now, and most of
> that crap is hard to maintain and disappears quick to the next-new-kid
> on the block.  ...
> Some new tool that will be dead
> next week makes sure that software dies.


I agree with your conclusions but not your premises <smile>. Many of  
the dvcs tools have been around for a long time (in software years),  
are used by zillions of people and work much better than CVS for lots  
and lots of use cases. OTOH, I no longer see the point of Subversion  
since it, at least, is not better enough than CVS.

If you do decide to move and do decide to use a DVCS, then I'd vote  
that we use git (I'm personally a big darcs fan but bit has much  
greater momentum, is "more" cross-platform and is building a bigger  
echo-system, ummm, ecosystem.) Other thoughts?

--
Gary Warren King, metabang.com
Cell: (413) 559 8738
Fax: (206) 338-4052
gwkkwg on Skype * garethsan on AIM * gwking on twitter

Re: [cclan-list] aside from the pangs of history, what about moving ASDF to common-lisp.net

[GP l. <fp...@cl...>]

   Date: Mon, 12 Jan 2009 20:57:00 -0500 (EST)
   From: Daniel Herring <dhe...@te...>

   > If not, then what do folks think about moving it to common-lisp.net?

   Sounds fine to me.

   Either way, I'd recommend joining the DVCS club:

   CVS was never fun

Actually, I'm rather tired of the current "but it's better" nonsense
on version control systems.  I've heard a dozen names now, and most of
that crap is hard to maintain and disappears quick to the next-new-kid
on the block.  I suppose that the illusion of having a 'real
up-to-date rvcs' makes up for a lack of progress elsewhere.

A old well-worn tool has little problem amongst the users of some
software, it makes it easy to obtain.  Some new tool that will be dead
next week makes sure that software dies.  If the purpose of software
improvement is to increase acceptance among the userbase, then making
the software easy to obtain is rather important.


[1] for instance, currently Subversion is broken here, since it
requires a version of Neon that breaks some important software.  Great
job!  Neon is apparently a piece of shit that has zero backwards
compatibility, and that ranks Subversion for me too.  Over the past
few years, it's gotten to the point that discovering some new software
package has yet-another-rcvs in use usually results dropping interest.

Re: [cclan-list] aside from the pangs of history, what about moving ASDF to common-lisp.net

[Daniel H. <dhe...@te...>]

On Mon, 12 Jan 2009, Gary King wrote:

> Are there good reasons to keep ASDF on sourceforge?

The email archive?
http://asdf.sourceforge.net/ ? ;)


> If not, then what do folks think about moving it to common-lisp.net?

Sounds fine to me.

Either way, I'd recommend joining the DVCS club:
http://arstechnica.com/news.ars/post/20090107-dvcs-adoption-is-soaring-among-open-source-projects.html

CVS was never fun; its much less so now that we have better tools.

- Daniel

P.S.  Is telent.net still operational?  It never works for me, but its 
advertised on http://cclan.sourceforge.net/

Re: [cclan-list] ASDF on MSWin

[Gary K. <gw...@me...>]

Hi Daniel,

I'm not sure why I dropped this by the wayside and will try to pick it  
up and toss it about this week.

regards,

On Jan 11, 2009, at 9:00 PM, Daniel Herring wrote:

> <PING>
>
> ASDF support on MSWin is patchy.  While CLisp treats MS shortcut  
> files as
> symlinks, many other lisps do not.  Since shortcuts don't work, some
> people end up adding sysdef-subdir-search to
> *system-definition-search-functions* (see cliki).
>
>
> In 2007, there was a patch for .lnk files; but it was never committed.
> https://sourceforge.net/mailarchive/forum.php?thread_name=87ir98pisj.fsf%40cantab.net&forum_name=cclan-list
>
> The patch (against 1.102) is still available at
> http://www.lichteblau.com/blubba/shortcut/
>
> It seems to apply cleanly to current ASDF; but my MSWin box isn't
> available for testing at the moment.
>
> Is there any chance of getting this vetted and into mainline ASDF?
>
> Thanks,
> Daniel
>
> ------------------------------------------------------------------------------
> Check out the new SourceForge.net Marketplace.
> It is the best place to buy or sell services for
> just about anything Open Source.
> http://p.sf.net/sfu/Xq1LFB
> _______________________________________________
> cclan-list mailing list
> ccl...@li...
> https://lists.sourceforge.net/lists/listinfo/cclan-list

--
Gary Warren King, metabang.com
Cell: (413) 559 8738
Fax: (206) 338-4052
gwkkwg on Skype * garethsan on AIM * gwking on twitter

[cclan-list] aside from the pangs of history, what about moving ASDF to common-lisp.net

[Gary K. <gw...@me...>]

Are there good reasons to keep ASDF on sourceforge?

If not, then what do folks think about moving it to common-lisp.net?

thanks,
--
Gary Warren King, metabang.com
Cell: (413) 559 8738
Fax: (206) 338-4052
gwkkwg on Skype * garethsan on AIM * gwking on twitter

[cclan-list] ASDF on MSWin

[Daniel H. <dhe...@te...>]

<PING>

ASDF support on MSWin is patchy.  While CLisp treats MS shortcut files as 
symlinks, many other lisps do not.  Since shortcuts don't work, some 
people end up adding sysdef-subdir-search to 
*system-definition-search-functions* (see cliki).


In 2007, there was a patch for .lnk files; but it was never committed.
https://sourceforge.net/mailarchive/forum.php?thread_name=87ir98pisj.fsf%40cantab.net&forum_name=cclan-list

The patch (against 1.102) is still available at
http://www.lichteblau.com/blubba/shortcut/

It seems to apply cleanly to current ASDF; but my MSWin box isn't 
available for testing at the moment.

Is there any chance of getting this vetted and into mainline ASDF?

Thanks,
Daniel

Re: [cclan-list] ASDF commit

[Nikodemus S. <nik...@ra...>]

On Sat, Jan 10, 2009 at 9:52 PM, Gary King <gw...@me...> wrote:
>> 2009-01-10  gw...@me...
>>
>>       added make-defined-systems-table (unexported)
>>       (re)applied Luis Oliveria's patch to add a restart for missing
>>       components. This patch works in SBCL but not in Allegro Common
>>       Lisp or CLISP. The problem has to do with the computation of
>>       restarts and I'm still trying to figure it out. Since the
>>       failure doesn't cause any _new_ problems (it just means that
>>       the putative new restart isn't listed), I'm still committing.
>>
>>       * asdf.lisp
>>
>>       tests for the above
>>
>>       (new) test/test-retry-loading-component-1.script
>>       (new) test/try-reloading-1.asd
>>       (new) test/try-reloading-dependency.hidden

Dunno what is the issue, but FWIW, I would structure this differently:
make the first form of RESTART-CASE an ERROR form, so the restart is
associated with the condition, and need for :TEST is eliminated. Not
sure if it would help with the issue or not.

Sketch below:

    (labels ((do-one-dep (required-op required-c required-v)
               (tagbody
                :retry
                  (let* ((dep-c (or (find-component
                                     (component-parent c)
                                     ;; XXX tacky.  really we should build the
                                     ;; in-order-to slot with canonicalized
                                     ;; names instead of coercing this late
                                     (coerce-name required-c) required-v)
                                    (go :error)))
                         (op (make-sub-operation c operation dep-c
required-op)))
                    (return-from do-one-dep (traverse op dep-c)))
                :error
                  (let ((e (if required-v
                               (make-condition 'missing-dependency-of-version
                                               :required-by c
                                               :version required-v
                                               :requires required-c)
                               (make-condition 'missing-dependency
                                               :required-by c
                                               :requires required-c))))
                    (restart-case
                        (error e)
                      (retry ()
                        :report (lambda (s)
                                  (format s "~@<Retry loading component ~S.~@:>"
                                          required-c))
                        (go :retry))))))
    ...

Cheers,

 -- Nikodemus

[cclan-list] ASDF commit

[Gary K. <gw...@me...>]

> 2009-01-10  gw...@me...
>
> 	added make-defined-systems-table (unexported)
> 	(re)applied Luis Oliveria's patch to add a restart for missing
> 	components. This patch works in SBCL but not in Allegro Common
> 	Lisp or CLISP. The problem has to do with the computation of
> 	restarts and I'm still trying to figure it out. Since the
> 	failure doesn't cause any _new_ problems (it just means that
> 	the putative new restart isn't listed), I'm still committing.
>
> 	* asdf.lisp
> 	
> 	tests for the above
> 	
> 	(new) test/test-retry-loading-component-1.script
> 	(new) test/try-reloading-1.asd
> 	(new) test/try-reloading-dependency.hidden
>


--
Gary Warren King, metabang.com
Cell: (413) 559 8738
Fax: (206) 338-4052
gwkkwg on Skype * garethsan on AIM * gwking on twitter

Re: [cclan-list] MISSING-DEPENDENCY not providing some retry restart?

[Gary K. <gw...@me...>]

I'm about to commit a fix for this that works in SBCL but not,  
strangely, ACL or CLISP. More details in the commit e-mail.


On Jan 10, 2009, at 10:27 AM, Tobias C. Rittweiler wrote:

>
> Is there a reason why there's no retry restart for MISSING- 
> DEPENDENCY in
> TRAVERSE? I'd like to invoke it from the Slime debugger after I
> installed the missing dependency.
>
>  -T.
>
>
> ------------------------------------------------------------------------------
> Check out the new SourceForge.net Marketplace.
> It is the best place to buy or sell services for
> just about anything Open Source.
> http://p.sf.net/sfu/Xq1LFB
> _______________________________________________
> cclan-list mailing list
> ccl...@li...
> https://lists.sourceforge.net/lists/listinfo/cclan-list

--
Gary Warren King, metabang.com
Cell: (413) 559 8738
Fax: (206) 338-4052
gwkkwg on Skype * garethsan on AIM * gwking on twitter

Re: [cclan-list] MISSING-DEPENDENCY not providing some retry restart?

[Gary K. <gw...@me...>]

Hi Tobias,

There is no reason I know of. As for Luis's patch, I think I remember  
applying that a long time ago. It may have been lost in some shuffle.  
I will re-apply it real soon now unless there is an unusual hue and cry.


On Jan 10, 2009, at 12:52 PM, Luis Oliveira wrote:

> "Tobias C. Rittweiler" <tc...@fr...> writes:
>
>> Is there a reason why there's no retry restart for MISSING- 
>> DEPENDENCY in
>> TRAVERSE? I'd like to invoke it from the Slime debugger after I
>> installed the missing dependency.
>
> Either because my patch sucks or because it's hard to get patches into
> ASDF. (Or both?)
>
> http://common-lisp.net/~loliveira/patches/asdf-component-load-retry-v4.patch
> http://thread.gmane.org/gmane.lisp.cclan.general/623
>
> -- 
> Luís Oliveira
> http://student.dei.uc.pt/~lmoliv/
>
>
> ------------------------------------------------------------------------------
> Check out the new SourceForge.net Marketplace.
> It is the best place to buy or sell services for
> just about anything Open Source.
> http://p.sf.net/sfu/Xq1LFB
> _______________________________________________
> cclan-list mailing list
> ccl...@li...
> https://lists.sourceforge.net/lists/listinfo/cclan-list

--
Gary Warren King, metabang.com
Cell: (413) 559 8738
Fax: (206) 338-4052
gwkkwg on Skype * garethsan on AIM * gwking on twitter

Re: [cclan-list] MISSING-DEPENDENCY not providing some retry restart?

[Luis O. <lu...@gm...>]

"Tobias C. Rittweiler" <tc...@fr...> writes:

> Is there a reason why there's no retry restart for MISSING-DEPENDENCY in
> TRAVERSE? I'd like to invoke it from the Slime debugger after I
> installed the missing dependency.

Either because my patch sucks or because it's hard to get patches into
ASDF. (Or both?)

http://common-lisp.net/~loliveira/patches/asdf-component-load-retry-v4.patch
http://thread.gmane.org/gmane.lisp.cclan.general/623

-- 
Luís Oliveira
http://student.dei.uc.pt/~lmoliv/

[cclan-list] MISSING-DEPENDENCY not providing some retry restart?

[Tobias C. R. <tc...@fr...>]

Is there a reason why there's no retry restart for MISSING-DEPENDENCY in
TRAVERSE? I'd like to invoke it from the Slime debugger after I
installed the missing dependency.

  -T.

Re: [cclan-list] Add CLEAN-OP to ASDF?

[Christophe R. <cs...@ca...>]

Robert Goldman <rpg...@si...> writes:

> This sounds like the right approach, and I will see if I can figure out
> how to modify Marco's code accordingly.  One question, though --- why
> should the :clean-for operation default to load-op instead of
> compile-op?  That was the one that seemed like the obvious choice to me.

You're quite right, it should be.

Best,

Christophe

Re: [cclan-list] Add CLEAN-OP to ASDF?

[Gary K. <gw...@me...>]

Hi Robert,

My two cents is that adding this would be a Good Thing.

Thanks for doing the work


On Dec 30, 2008, at 10:55 AM, Robert Goldman wrote:

> Christophe Rhodes wrote:
>> Robert Goldman <rpg...@si...> writes:
>>
>>> E.M. Baringer's Arnesi library has an ASDF clean-op in its src/ 
>>> asdf.lisp
>>> file (thanks to fe[nl]ix on #lisp for the pointer).  Is there any  
>>> reason
>>> this operation could not be imported into the core ASDF  
>>> distribution?
>>> Cleaning out old FASLs seems like a very nice thing to have.
>>
>> The usual argument against is that clean-op is philosophically wrong:
>> the output-files of which operation should it clean?
>>
>> I'm no longer sure that that argument holds water: the operation
>> itself can know, by virtue of referring to the operation it's  
>> cleaning
>> up for.  Something like
>>
>> (defclass clean-op (operation)
>>  ((clean-for :initarg :clean-for))
>>  (:default-initargs :clean-for (make-instance 'load-op)))
>>
>> Now (asdf:oos 'asdf:clean-op ...) will do the right thing, and those
>> with special requirements can additionally do things like
>>  (asdf:oos 'asdf:clean-op ... :clean-for (make-instance 'test-op))
>> or similar.
>>
>> This probably isn't going to be a straightforward import of Marco's
>> code (though maybe it is, I don't know); on the other hand, if this
>> functionality were available in this form I would be happy.
>>
>> Best,
>>
>> Christophe
>>
>> ------------------------------------------------------------------------------
>> _______________________________________________
>> cclan-list mailing list
>> ccl...@li...
>> https://lists.sourceforge.net/lists/listinfo/cclan-list
>>
>
> I should have reread Marco's code more carefully before responding.   
> His
> clean-op already takes a :for-op initarg that does what Christophe
> suggests here.  I didn't give him sufficient credit.
>
> But the default is compile-op, not load-op.
>
> I'm going to start by splicing the clean-op into my local copy of  
> asdf,
> and will propose a patch if testing proves successful.
>
> Oh, yes, and I have written to Marco to ask if inserting his code  
> would
> be acceptable, and under what circumstances.  In particular, Arnesi  
> was
> released under the Open Source license, whereas ASDF is released under
> MIT/X license terms, and Marco asks for his copyright notice to be
> preserved.
>
> Best,
> R
>
> ------------------------------------------------------------------------------
> _______________________________________________
> cclan-list mailing list
> ccl...@li...
> https://lists.sourceforge.net/lists/listinfo/cclan-list

--
Gary Warren King, metabang.com
Cell: (413) 559 8738
Fax: (206) 338-4052
gwkkwg on Skype * garethsan on AIM

Re: [cclan-list] Add CLEAN-OP to ASDF?

[Robert G. <rpg...@si...>]

Christophe Rhodes wrote:
> Robert Goldman <rpg...@si...> writes:
> 
>> E.M. Baringer's Arnesi library has an ASDF clean-op in its src/asdf.lisp
>> file (thanks to fe[nl]ix on #lisp for the pointer).  Is there any reason
>> this operation could not be imported into the core ASDF distribution?
>> Cleaning out old FASLs seems like a very nice thing to have.
> 
> The usual argument against is that clean-op is philosophically wrong:
> the output-files of which operation should it clean?
> 
> I'm no longer sure that that argument holds water: the operation
> itself can know, by virtue of referring to the operation it's cleaning
> up for.  Something like
> 
> (defclass clean-op (operation)
>   ((clean-for :initarg :clean-for))
>   (:default-initargs :clean-for (make-instance 'load-op)))
> 
> Now (asdf:oos 'asdf:clean-op ...) will do the right thing, and those
> with special requirements can additionally do things like
>   (asdf:oos 'asdf:clean-op ... :clean-for (make-instance 'test-op))
> or similar.
> 
> This probably isn't going to be a straightforward import of Marco's
> code (though maybe it is, I don't know); on the other hand, if this
> functionality were available in this form I would be happy.
> 
> Best,
> 
> Christophe
> 
> ------------------------------------------------------------------------------
> _______________________________________________
> cclan-list mailing list
> ccl...@li...
> https://lists.sourceforge.net/lists/listinfo/cclan-list
> 

I should have reread Marco's code more carefully before responding.  His
clean-op already takes a :for-op initarg that does what Christophe
suggests here.  I didn't give him sufficient credit.

But the default is compile-op, not load-op.

I'm going to start by splicing the clean-op into my local copy of asdf,
and will propose a patch if testing proves successful.

Oh, yes, and I have written to Marco to ask if inserting his code would
be acceptable, and under what circumstances.  In particular, Arnesi was
released under the Open Source license, whereas ASDF is released under
MIT/X license terms, and Marco asks for his copyright notice to be
preserved.

Best,
R

Re: [cclan-list] Add CLEAN-OP to ASDF?

[Robert G. <rpg...@si...>]

Christophe Rhodes wrote:
> Robert Goldman <rpg...@si...> writes:
> 
>> E.M. Baringer's Arnesi library has an ASDF clean-op in its src/asdf.lisp
>> file (thanks to fe[nl]ix on #lisp for the pointer).  Is there any reason
>> this operation could not be imported into the core ASDF distribution?
>> Cleaning out old FASLs seems like a very nice thing to have.
> 
> The usual argument against is that clean-op is philosophically wrong:
> the output-files of which operation should it clean?
> 
> I'm no longer sure that that argument holds water: the operation
> itself can know, by virtue of referring to the operation it's cleaning
> up for.  Something like
> 
> (defclass clean-op (operation)
>   ((clean-for :initarg :clean-for))
>   (:default-initargs :clean-for (make-instance 'load-op)))
> 
> Now (asdf:oos 'asdf:clean-op ...) will do the right thing, and those
> with special requirements can additionally do things like
>   (asdf:oos 'asdf:clean-op ... :clean-for (make-instance 'test-op))
> or similar.
> 
> This probably isn't going to be a straightforward import of Marco's
> code (though maybe it is, I don't know); on the other hand, if this
> functionality were available in this form I would be happy.
> 
> Best,
> 
> Christophe
> 

This sounds like the right approach, and I will see if I can figure out
how to modify Marco's code accordingly.  One question, though --- why
should the :clean-for operation default to load-op instead of
compile-op?  That was the one that seemed like the obvious choice to me.


Best,
r

Re: [cclan-list] Add CLEAN-OP to ASDF?

[Christophe R. <cs...@ca...>]

Robert Goldman <rpg...@si...> writes:

> E.M. Baringer's Arnesi library has an ASDF clean-op in its src/asdf.lisp
> file (thanks to fe[nl]ix on #lisp for the pointer).  Is there any reason
> this operation could not be imported into the core ASDF distribution?
> Cleaning out old FASLs seems like a very nice thing to have.

The usual argument against is that clean-op is philosophically wrong:
the output-files of which operation should it clean?

I'm no longer sure that that argument holds water: the operation
itself can know, by virtue of referring to the operation it's cleaning
up for.  Something like

(defclass clean-op (operation)
  ((clean-for :initarg :clean-for))
  (:default-initargs :clean-for (make-instance 'load-op)))

Now (asdf:oos 'asdf:clean-op ...) will do the right thing, and those
with special requirements can additionally do things like
  (asdf:oos 'asdf:clean-op ... :clean-for (make-instance 'test-op))
or similar.

This probably isn't going to be a straightforward import of Marco's
code (though maybe it is, I don't know); on the other hand, if this
functionality were available in this form I would be happy.

Best,

Christophe

[cclan-list] Add CLEAN-OP to ASDF?

[Robert G. <rpg...@si...>]

E.M. Baringer's Arnesi library has an ASDF clean-op in its src/asdf.lisp
file (thanks to fe[nl]ix on #lisp for the pointer).  Is there any reason
this operation could not be imported into the core ASDF distribution?
Cleaning out old FASLs seems like a very nice thing to have.

I don't *believe* that the license terms would cause any problems; we
would just have to give credit to Marco.  But IANAL.

Best,
Robert