[ltp] More war stories about Lenovo build quality

[David A. Desrosiers]

> Stupid question --- you have used the Windows Task Manager to make 
> sure you don't have something burning huge amounts of CPU time in 
> the background, right?

Yes, and it's lean and light in Windows, not much eating CPU except 
maybe ZoneAlarm.

> Um, how are you configuring thinkfinger?  It should have access to 
> your typed password at all?  I have it set up to bypass password 
> checking if successful, but it doesn't store a copy of my password 
> at all.

That's the problem: When the fingerprint scans fail, as they do 80% of 
the time or more when the reader overheats, it falls back to password 
authentication. The problem is that there is some "random" delay that 
thinkfinger uses when it recycles after a failure (roughly 5-9 
seconds, on my T61p). If I don't wait long enough, I end up typing my 
password out and it's echoed to the console.

After a failure, there's an extra carriage return, then a "Sorry, try 
again." prompt, then another delay, and then the "Password or swipe 
finger:" prompt. It is in-between these last two bits where I have to 
be careful not to type my password too fast. It's minor, but it's 
annoying, because the timeout seems to be random.

> At least with Ubuntu Gutsy and running a bleeding edge kernel (on an 
> X61s with an Intel graphics chipset), I haven't had any issues....

I think I've nailed this down a bit further. When the system "thinks" 
it is under heavy load, the keyboard driver goes nuts, and forgets its 
meta keys, layout and begins duplicating keys hundreds of times.

Unfortunately, the load is unavoidable because my entire drive is an 
encrypted LVM, so 30% of my drive performance and IO is eaten for 
kcryptd right off the top. When I load up a vmware session that comes 
from that drive, my load is normally 13 or higher. If I move that VM 
to the unencrypted drive in my UltraBay slot, the load remains under 
2.

There is definitely a direct relationship to system load and the flaky 
input devices. Lots of Ubuntu bugs have been filed against the 
keyboard repeat bug, but nobody has really nailed it down yet. If you 
create an artificial load on the system, the problem magnifies itself 
exponentially.

> I use crossover office and the Intel drivers on Ubuntu Gutsy and I 
> haven't noticed any problems.

I had 3 people in #ubuntu+1 verify this a few nights ago. Try this: 
Install SafariSetup.exe with wine, then after it is installed, launch 
Safari and go to Edit -> Preferences. Bam. Every single time I do 
that, X crashes out and gdm recycles. The people I had testing it 
either had X crash, or wine completely locked up right when they went 
to the Edit menu. It's 100% reproducable.

IMHO, a bad app in wine should never pull the rug out from under X.

> (For example, I very much wonder about the decision of replacing FF2 
> with FF3, given some of the usability issues I've heard reported 
> regarding FF3.  But that alone has made me hesitant to upgrade to 
> Hardy except on a crash and burn box --- and some of your reported 
> difficulties are also reasons for concern.)

I've been using Debian and Ubuntu long enough to know which things to 
accept as repo packages, and which things to gut out and run on my own 
from upstream packages. Firefox, OpenOffice.org, kernels, Java and 
about a dozen other things all come from upstream sources, not the 
.deb repositories, because of that reason.

FF3 is a mess of a GUI, and I reported it to the folks in the proper 
channels, including screenshots of the gross UI layout and bugs. Sure, 
it's beta, but it is so vastly different from 2.x that I probably will 
never make the jump.

Thanks for the tips. I'm not giving up on Linux just yet.