Re: [NTLK] Fun with new CF card and MP3's

From: Paul Guyot (pguyot_at_kallisys.net)
Date: Wed Nov 26 2003 - 15:57:35 PST


Aux environs du 26/11/03 à 13:54 -0500, sous le titre "[NTLK] Fun
with new CF card and MP3's", Dave Bonhoff prit sa plus belle plume
pour écrire les mots suivants:
>The first thing that I noticed is that Heaven or Hell won't run if any
>partition of the card is mounted. I had it set to run when powering
>down and to show the progress bar. It pops up and just as quickly
>disappears before my Newt shuts down. Has anyone else noticed this
>behavior?

It's a bug in Heaven or Hell. Move a package to the partition (and
move it back if you don't want it to stay on the partition) and it
will work.

The bug is there is no packages soup on new ATA cards. Another
related bug is that NewtWorks may fail to open under some
circumstances like this, although I don't recall exact details. I
noticed this at the very beginning of ATA Support development, since
all my linear cards were not empty while my ATA partitions were. I
think I sent an e-mail to the developer of Heaven or Hell (and that
the NewtsWork bug is documented somewhere).

>Storage
>Sorry, a problem has occurred (-10622)

This is bad. I actually don't even have an idea of how you came up
with that error. It's object too big for store. I thought ATA Support
had no limit but in fact, I coded one (which has no reason to exist,
except that I thought linear stores had a limit of 64 KB - 1 byte and
therefore I've put that).

>I also have been restarting my Newt very frequently since it is hanging
>quite often, like while filing large files. Patience, I know, but
>sometimes I need to use my friend. Thank you, Paul, for journaling. I
>haven't lost any data or seen any corruption despite my actions! Do I
>need to change the cache sizes in the ATA prefs? If so, any
>suggestions what size?

Actually, I don't think I would have implemented journalizing in ATA
Support wouldn't the system require me to do so. The reason is that
it is what eats most of ATA Support resources. It's not like
journalizing on file systems, everything is in transactions, data
*and* metadata (vs just metadata on journalized file systems).
Increasing the caches helps. Since you have a MP2100, 30 or 50 pages
should improve things. I don't recommend these values for a MP2000
because it would use too many memory pages and slow down the Newton
by requiring more disk/card accesses to swap pages.

Also, the -10622 is really bad. Make sure the card behaves properly
indeed. I don't have any test suite that reveals data corruption even
if these test suites include very intensive storage operations
(including crashes and the such).
If you can reproduce the -10622 deterministically, and if you want, I
can cook a new build without this 64 KB limitation for you to try. As
I wrote, this limit only exists because I think it's what the system
has for linear cards/internal memory. The ATA Support data format has
no real limit on object sizes except the size of the store itself
(plus space to do the transaction). I'm not sure ATA Support actually
triggers this error because the system realistically asks for a
creation of an object that big. It might be that some data is
corrupted somehow and the system wants an object with an arbitrary
random size (much, much bigger than 64 KB).

Paul

-- 
Philosophie de baignoire - consultations sur rendez-vous.
NPDS/NewtonOS: http://newton.kallisys.net:8080/
Apache/FreeBSD: http://www.kallisys.com/
-- 
This is the NewtonTalk list - http://www.newtontalk.net/ for all inquiries
List FAQ/Etiquette/Terms: http://www.newtontalk.net/faq.html
Official Newton FAQ: http://www.chuma.org/newton/faq/


This archive was generated by hypermail 2.1.5 : Wed Nov 26 2003 - 19:30:01 PST