From ktilton at nyc.rr.com  Wed Feb 22 04:52:27 2006
From: ktilton at nyc.rr.com (Kenny Tilton)
Date: Tue, 21 Feb 2006 23:52:27 -0500
Subject: [cells-devel] Cells III Call for Abuse <g>
Message-ID: <43FBEE0B.8080801@nyc.rr.com>

Thomas Burdick has been pushing the envelope on Cells lately, and it 
turns out to have been A Good Thing. I am not sure we can call this 
Cells III, but I hate those namby-pambies who go up by hundredths of a 
version, so what the heck. :) And as I fix up the stuff Thomas managed 
to break, turns out the code is getting even simpler and more efficient. 
Always nice.

I do not expect much response -- one really has to have thought about 
Cells internals -- but it cannot hurt to ask. The deal is this:

I am anxious to have any test code folks think will break Cells. 
Especially challenging is a def-c-output SETF of a c-input mediated 
Cell, or a dependent Cell that recalculates its KIDS slot, or a 
combination wherein the output setf triggers a kids recalc /and/ the new 
kids have cells with rules that look for each other by name and then use 
ruled slots of each other. <gasp>

Anyway, this is my problem, just thought I would see if there is someone 
out there who takes diabolical pleasure from breaking code. :)

Peace, ken




From peter.denno at nist.gov  Wed Feb 22 17:46:01 2006
From: peter.denno at nist.gov (Peter Denno)
Date: Wed, 22 Feb 2006 12:46:01 -0500
Subject: [cells-devel] Cells III Call for Abuse <g>
In-Reply-To: <43FBEE0B.8080801@nyc.rr.com>
References: <43FBEE0B.8080801@nyc.rr.com>
Message-ID: <200602221246.01757.peter.denno@nist.gov>

Kenny,

I'll give it a try in the context of cells-gtk. Perhaps not for a week or so 
though. Is this CVS HEAD on the common-lisp.net repository?


On Tuesday 21 February 2006 23:52, Kenny Tilton wrote:
> Thomas Burdick has been pushing the envelope on Cells lately, and it
> turns out to have been A Good Thing. I am not sure we can call this
> Cells III, but I hate those namby-pambies who go up by hundredths of a
> version, so what the heck. :) And as I fix up the stuff Thomas managed
> to break, turns out the code is getting even simpler and more efficient.
> Always nice.
>
> I do not expect much response -- one really has to have thought about
> Cells internals -- but it cannot hurt to ask. The deal is this:
>
> I am anxious to have any test code folks think will break Cells.
> Especially challenging is a def-c-output SETF of a c-input mediated
> Cell, or a dependent Cell that recalculates its KIDS slot, or a
> combination wherein the output setf triggers a kids recalc /and/ the new
> kids have cells with rules that look for each other by name and then use
> ruled slots of each other. <gasp>
>
> Anyway, this is my problem, just thought I would see if there is someone
> out there who takes diabolical pleasure from breaking code. :)
>
> Peace, ken
>
>
> _______________________________________________
> cells-devel site list
> cells-devel at common-lisp.net
> http://common-lisp.net/mailman/listinfo/cells-devel

-- 
- Best regards,
   Peter


From ktilton at nyc.rr.com  Wed Feb 22 18:32:25 2006
From: ktilton at nyc.rr.com (Kenny Tilton)
Date: Wed, 22 Feb 2006 13:32:25 -0500
Subject: [cells-devel] Cells III Call for Abuse <g>
In-Reply-To: <200602221246.01757.peter.denno@nist.gov>
References: <43FBEE0B.8080801@nyc.rr.com>
	<200602221246.01757.peter.denno@nist.gov>
Message-ID: <43FCAE39.1050304@nyc.rr.com>

Peter Denno wrote:

>Kenny,
>
>I'll give it a try in the context of cells-gtk. Perhaps not for a week or so 
>though. Is this CVS HEAD on the common-lisp.net repository?
>  
>
Ah, no, but you make a good point. I was thinking folks like Thomas 
could just submit stuff that they think is nasty. But of course you will 
want to try it out and enhance it if it fails to break things. :)

If you want to wait, it should only be a few days... nahh, I am moving 
within the next five days, so add a two day delay for that. I might 
email you a distro rather than commit to CVS, unless I think it looks 
very solid.

Cells-GTk will be a good test. I have moved to Tk (atop LTk, but just 
using its core) to get Aqua on OS X, and i have discovered that Tk is a 
little sensitive to the order in which things get done. So I am 
extending DEF-C-OUTPUT to give the user the ability to queue up actions 
(such as talking to Tk) during any propagation and then at the end 
administer them with a user callback. For Tk I will just sort the 
actions to suit Tk.

GTk might not have the same problem, but if you have run into anything 
help is on the way.

Thx for pitching in.

kt



From ktilton at nyc.rr.com  Thu Feb 23 06:40:52 2006
From: ktilton at nyc.rr.com (Kenny Tilton)
Date: Thu, 23 Feb 2006 01:40:52 -0500
Subject: [cells-devel] Never mind...
Message-ID: <43FD58F4.8060203@nyc.rr.com>

I got so lost in the permutations that I think I will take a couple of 
days to list them all and then carefully construct a comprehensive test 
suite just to help me think about the rewrite. This will include 
attempted violations of the Cells discipline, such as setfing an input 
cell within an output method without wrapping it in a suitable macro 
that will defer the setf until propagation to dependencies is complete. 
This is a new discipline by the way. Really simplifies the code.

Thx anyway, kt