Hi all,
I'm pleased to announce ATA Support 1.0b9. I should say that this
release is there mostly because previous release has expired. I'm in
the process of making important modifications which can't be shipped
yet, so expect a new version later this week.
Anyway, 1.0b9 expires on November, 5th.
What's new?
-----------
* Fixed a bug (with a quick workaround) caused by the expiration date
mechanism when a card was insterted at startup.
Here's the story: I was getting the date with a system function which
happens to use NewtonScript objects (to get the timezone basically).
And this, from the card server task, so bad things happened.
Now, I just catch bad things and consider that the software didn't
expire. This is a dirty fix, but I prefer to spend time on other bugs
as the expiration date will disapear in the final release (also, it's
there only to force me to release new versions and be sure that tests
results I get are from the very latest version of the software).
* Fixed a bug in the drivers when writing to the cards which don't
use the default transfer width. Since 1.0b3, the software supports
various transfer width, but I forgot to make the required changes for
writing operations.
* Added informations about the card in the card infos, specifically
the transfer width and the ATA mode (the only ATA mode currently
implemented is the memory mode, but as soon as the driver will be in
a separated task, it will be very interesting to use interrupt-based
ATA modes).
* The documentation, although not finished (sorry Ben) is now included.
Ben: you can't backup without a third party software for the moment
(and this will very probably not be implemented for 1.0). OK, it
doesn't explain anything, please give me the time to finish to write
it.
* Nick's Test It takes on my Hitachi card 20.1-20.4 seconds (just
like with 1.0b8) to create 355 entries with a 400 bytes string in
each (in a nutshell, the transfer rate it mentions doesn't mean
anything). I'm writing this here because I used to rely on Carsten's
tests to know if I succeeded in optimizing the driver. BTW, thanks
Carsten.
I also added this to the documentation because I've been making tests
to see if what I thought to be an improvement (disabling
interruptions) helped. Apparently, it doesn't, so back to the
previous code.
I also tried to change the timing settings (they are 35% too high on
MP2x00/eMates because of possible accelerated units) and it didn't
help neither. So I think that I should work on asynchronous
operations and optimizing the transational persistent object store.
* I've added a thanks to every beta testers in the about box. I
really got a lot of help from a lot of users. Thanks to everybody.
* Implemented support to convert stores from one format to another, including:
- Alerts to tell user if a conversion operation occurred or failed.
- Format number in the store info slip (right bottom corner).
- Functions to retrieve and propagate this information to the NS interface.
- Various tests here and there to do the actual conversion.
* New format with a pool of 256 sectors (128 KB) allocated at the
end. These sectors can only be used when doing transactions
(equivalent to the slop in linear stores). There are some differences
with linear stores arithmetics, though. Note: this isn't sufficient
yet to delete a big object (typically a package) which is larger than
what's left in the store. This is a known limitation and should be
fixed soon.
* New arithmetics of ATA Stores. The current used size is the actual
used size, plus the pool (I think that the slop on linear stores
isn't included in the used size nor total size).
Known bugs:
-----------
* Some users have reported that when there are some packages (such as
NetHopper) on an ATA store, bad things happen when the Newton goes to
sleep and this causes the system to restart. I never experienced this
which may have been fixed by the new power-management code.
* Under some circumstances, the Newton is locked. This includes
situations when the 3Com's driver is installed or when one installs
packages such as NetHopper. I've been able to reproduce the latter,
and the Newton loops into a system function
(TFrameRelocationGenerator::Update) for an unknown reason.
* Some cards cannot be written to by the driver. This is why some
users experienced problems when partitionning.
* The eMate problem which restarts when they're closed hasn't been fixed.
* Some modifications are being made to the store management, even if
the current release should be rather stable.
Where to download it?
---------------------
Well, as usual, on my website.
http://www.kallisys.com/newton/ata/
You'll find the (partially written) documentation inline as well as
in the archives. Please note that the log version isn't included this
time.
Paul
-- Home page: http://www.kallisys.com/ Newton-powered WebServer: http://newt.dyndns.org:8080/-- This is the Newtontalk mailinglist - http://www.newtontalk.net To unsubscribe or manage: visit the above link or mailto:newtontalk-request_at_newtontalk.net?Subject=unsubscribe
This archive was generated by hypermail 2.1.2 : Thu Nov 01 2001 - 10:02:26 EST