[vdr] Re: dvb threatening file system? [was Drop support for multiple video dirs?]

[Guido Fiala <gfiala@s.netic.de>]

Am Dienstag, 7. Oktober 2003 14:38 schrieben Sie:
> How do you think will such a test look like? You cannot test all components
> in all their different constellations. That's the reason, why systems with

Of course not, but running all programs to be used at the system at once and 
in action is not a bad test...

> high stability requirements make use of "industrial computers" which are
> well combined and tested components. If you puzzle your own system

Of course - but is there one that fits all your needs and is not to expensive?

> and even make use of noname components like memory, power supply, etc. you
> cannot expect, that this hardware will run in a stable way.

Right - but all the components (beside RAM and CPU) are onboard-components of 
an i815-mainboard - they should work together...

> You cannot compare (for example) a little text editor with a hw-related
> driver and sw, or do you? If your hw isn't stable it's quite sure that most
> of the uncritical sw will run fine. But if you want to run hw-related sw
> you may expect such a behaviour.

The comparison is not intended - it was just meant as a warning not to trust 
simple burn-in tests (especially "real-mode" or tests that do not include all 
the cpu's instructions). 
I bought lots of systems, even for the office from named providers, they said 
they run burn-in tests. I did too (unter windows). Memtest did never found an 
error - and yet, the system was unstable, discovered it by several subsequent 
kernel-compiles after some applications did crash that never crashed before. 
In that cases i had to run the AMD-processor at 1000 instead of 1500 to make 
it stable.

As one can not expect an hardware seller (did you already tried to ask for in 
your favorite computer store?) to run such extensive tests a self-mad system 
from components is'nt all that bad - if one does'nt work one can always give 
it back...

> I definitly can confirm that memory chips from noname-manufacturer causes
> unpredictable behaviour of a system. If you use in the same system an
> infineon-memory you won't have any problems.

If your read carefully you see that i have indeed tested only Infineon's !
An name alone is not a stone...

> I cannot confirm that. If you buy a 133Mhz memory, this chip have to run at
> 133Mhz or it is defect. But some of the noname-manufacturer think that an

You are right - but the specification of "running" is that varies...



-- 
Info:
To unsubscribe send a mail to ecartis@linuxtv.org with "unsubscribe vdr" as subject.