| 1 |
On Wednesday 31 January 2007 14:27, Dave Oxley wrote: |
| 2 |
> Hi, |
| 3 |
> |
| 4 |
> I have a MythTV frontend running Gentoo on an AMD 64bit with an |
| 5 |
> Nvidia 6600 graphics card. I have had no end of problems with the |
| 6 |
> graphics card (Xid errors). It crashes with a white line across the |
| 7 |
> screen regularly when 3d goom is up when I'm playing music. I've so |
| 8 |
> far been putting it down to the nvidia driver but I'm beginning to |
| 9 |
> think I may have a dodgy stick of RAM or something as I also get |
| 10 |
> Segmentation faults when doing emerge's on occasion. When this |
| 11 |
> happens it says its an OS error and retrying the emerge normally |
| 12 |
> fixes it. How can I find out what the problem is? Is there a way of |
| 13 |
> testing the RAM? Is it possible that it is the nvidia drivers still |
| 14 |
> (I am often playing music when I run an emerge)?? |
| 15 |
|
| 16 |
Faulty or dodgy RAM is indeed often the cause behind random segfaults. |
| 17 |
The best test of all is gcc which is why you get the error more with |
| 18 |
emerge than anything else. gcc just happens to stress the ram and disk |
| 19 |
system in ways that few other peices of software do. |
| 20 |
|
| 21 |
There is a product called memtest86 which tests ram, but you have to run |
| 22 |
it for many many hours to be sure of ewhat it's saying. And, I have |
| 23 |
read reports that what memtest86 does is predictable, whereas a big |
| 24 |
emerge is much more random. |
| 25 |
|
| 26 |
How to test: Well, you could come up with a $40k once-off rig to test |
| 27 |
memory that costs a fdraction of that. The easy way is to just swap out |
| 28 |
the memory and try again. Buy some, it's not that expensive and you |
| 29 |
*will* use it somewhere soon anyway :-) Or borrow some DIMMS from a |
| 30 |
friend for a day or three |
| 31 |
|
| 32 |
alan |
| 33 |
|
| 34 |
-- |
| 35 |
gentoo-user@g.o mailing list |
Archives