From: Paul Guyot (pguyot_at_kallisys.net)
Date: Tue Jul 15 2003 - 23:16:11 PDT
Il me semble que le 15/07/03 à 19:57 -0700, Richard David Turner, SG
nous racontait:
>That's interesting. I'm using a 256MB ATA CF card with Paul's driver
>that at I've currently divided into 3 stores of 62MB, and one of
>62.5MB. At first, the only problems I've seen, so far, is that I can
>not make just ONE store from the whole card. Nor can I make 2 stores of
>maximum/equal size (124MB). But I can make one SMALL 4MB store AND two
>maximum stores of about 120MB-124MB each. That SEEMS to work alright.
OK. David, I've started writing this new partition interface but it
takes some time and I should better work on my thesis. I'll send you
the thing at the end of the next week.
I'll also get a 256+ card. Since a lot of people (so far, 4, maybe 5
if I count Ed) have this problem, I'll be able to reproduce it and
hopefully this won't take long to fix. I already have half dozen of
fixes that haven't been released.
I don't think (although it is a possibility) the problem comes from
the system, i.e. the system doesn't like stores bigger than 122/123
MB. Since normally the system doesn't know about the actual size of
the store, I just handle objects for it.
>Though within the past couple of weeks, I have noticed some sort of
>'corruption' of the 62.5MB store(the 4th) whenever I mount the just the
>4th store by itself, or together with either the 3rd or 2nd. I'll
>sometimes lose access to the Routing(Envelope) menus as soon as the 4th
>store is mounted, or after any other store is mounted AFTER the 4th.
>But ONLY if the 4th store is mounted. If any combo of the other three
>is mounted-No problem.
My past experience with ATA Support is that, from user records, it's
not perfectly as stable as linear stores (which aren't perfect
either). My NPDS server runs official RC5 build and doesn't have any
problem, although I know there are issues with InkSpot and Newtscape.
Many users never had any problem, apparently.
But a lot of reports I got were actually caused by software on the
stores. Newton software are supposed to be easily deactivated, but
very few are properly cleaning the world when they're leaving.
BTW, I've finally found out why package installation takes so long.
There is a design problem here, I discovered this updating the
internal specs of ATA Support for Philz. Basically, while for regular
transactions I work on the original copy and I discard the backup
when the transaction is committed, with separated transactions, I
work on the backup and I copy it back when the transaction is
committed. I don't recall why I made this design choice, there is
surely a reason (and the other way around would probably use more
space on the store during a transaction). I just don't want to make
it right before being able to dive completely into the code.
Something linear stores are pretty good at is transactions because
they just switch a bit to validate them.
I also noticed a possible (yet to be checked) leak on the store if a
crash happens in a given state.
Paul
-- Philosophie de baignoire - consultations sur rendez-vous. NPDS: http://newton.kallisys.net:8080/ Apache: 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 : Tue Jul 15 2003 - 23:30:01 PDT