Re: [Chicken-users] Chicken Gazette - Issue 13

[F. Wittenberger]

Hi Chickeners!

We need a new mailing list.

No, wait.  I don't want to follow this on any mailing list.  Please
let's have an RSS feed for those things I will all too often have hit
the button with the label "mark all as read" for live time reasons.

So.  All following belongs into the section "letters from our readers",
where all those famous dog owner's excuses go (like "what do are you
asking for??? -- a dog is a valuable - good? like a tree - good? so it's
the same as with trees, they are valuables - good? so when the apple
falls down at your side of the fence, it's your apple - good?  so when
the sh* falls down in your garden it's not my dog's anymore - good?  --
So you can enjoy to have an old dog like mine who learned that having a
fence close to the ass makes a difference when it comes to clean
yourself up afterwards ;-).

Still with me?  Hit DEL now!




You have been warned!
^L



Am Montag, den 22.11.2010, 21:22 +0100 schrieb Peter Bex:
> == 2. Core development
> The scrutinizer was updated to give a warning when a one-armed `if`
> is used in tail-position, as suggested on chicken-users by Jörg
> Wittenberg.

sure?  or is -- no, wait, no "git" struggle now.  spare me, please!

> A new `equal=?` procedure was added, which works like `equal?` except
> that it ignores exactness when comparing numbers.

So, which version is NOT R5RS compliant?

> == 3. Chicken Talk
> 
> Just before the previous gazette was issued, Nicolas Pelletier posted
> a follow-up to an old thread of his about an overflow problem with the
> `current-time` procedure.  He reported that he had tested the changes
> Felix made in the "flonum-milliseconds" branch, and they were working
> perfectly for him.

This is something, I'd really like to put some maybe relevant remarks
in, which could easily turn out to be irrelevant fears of mine.

I'm currently running my stuff on a "slightly" modified scheduler
implementation.  It uses left leaning read black trees instead of flat
lists for file descriptors and timed events.  Scales kinda better, you
know.  Apart of that, but starting from there, I found some threading
issues which I don't recall in detail right now.

So far following the "flonum-milliseconds" branch for me will require to
change even more in the scheduler.scm first (replace <int-llrbtree...>
with <flownum-llrbtree...>) and then test this.  The latter may or may
not be the "irrelevant fears of mine":

My test case is an askemos network.  A mixture of 32bit/64bit Intel on
Linux/FreeBSD where the Askemos/BALL executable is compiled via either
rscheme or chicken plus an Sheeva Plug with Linux and chicken only
(rscheme does not compile there).

So to get a single transaction through, it's necessary that all those
combinations will compute the very same checksum for the same input.

The fear is, that by changing the arithmetic type of the time, it might
happen, and even worse: might rarely or only after some flownum
implementation dependant point in time happen, that we end up with
different time values at some of the nodes.  I.e., end of processing.

Therefore I currently prepare my code for a compile time switch to use
either flownum or fixnum for milliseconds in the hope to eventually
integrate again with the mainline.

...takes time...

> The latter half of the week saw a flurry of messages on the list:
> 
> On Wednesday, Jörg Wittenberg posted some ideas about how to improve
> Chicken's debugging information so that it would be able to detect a
> difficult problem he had encountered (but already solved).

Now recap the dog owner's excuse!

I'm glad to learn that I solve the problem!

Too much credit actually.  The problem disappeared magically.

And that's the problem I have: I don't (fully) trust any magic.

> With this week's omelette recipe we'll dip our toes in some of the
> more advanced uses of the insanely great PostgreSQL database. With the
> postgresql egg I've been trying to unlock some advanced features
> through a high-level API, so it becomes a breeze to use them!

Now, wait!

As I said with a smiley in the other posting: I would never enter into a
religious war.

Therefore I have to raise a flag in favour of the great SQLite data
base!

Right here!

Why?:  How does your PostgreSQL handle master-master replication?  What
if I mount a man in the middle attack on one of your master replicas and
inject fake update packets?  Will I be able to tamper with your data
base?  Too bad for you!!!  ;-)

</end-flamewar ;->

These day's I commented a straight forward application, which I want to
convert into a beginners tutorial.  It intermixes some "magic xml", XSL,
Scheme and SQL ending up in something I (biased as an author can be)
would call "concise" (web)application.  (Those who can read/translate
those few German words in the actual user interface might find even find
it funny when contrasted to the English source code comments.)

Since the thing runs on chicken on this geeky self made 5 core ARM
"laptop": http://www.askemos.org/A0babc647c6560d9c2e19d26377830c04
(as well as the above mentioned heterogeneous network)
I consider it hereby on topic and ask everyone to take a minute to skim
over 
http://www.askemos.org/Adc5dd0c30f6e63932811ed60e019bb2d/Kalender?date=2010-11-01
and ask any questions you encounter.  I'd really like to find out what
needs to be documented first/better, where I'm introducing too many
things at once etc.

Thanks a lot

/Jörg