Curse those computers!

[IBBoard]

I'm definitely a techno-geek, albeit one without enough money to indulge in gadgets. That means I like tinkering and I like Linux, and it also means that I can generally fix things when they go wrong.

Recently, the wife's Windows-based desktop machine has been blue-screening at random times (sometimes in a game, sometimes while running iTunes, sometimes seemingly on shutdown, because it complains it didn't shut down cleanly even though it seemed to. After I was eventually allowed at it I started with a memtest, because a disk check had come back clean and bad memory would be easier to diagnose than a bad power supply, and one stick had a bad chip on it. It's probably the first time I remember it hapening to me, but the computer is about four years old. It's now on only 1GB of memory, but hopefully it'll keep going because it is a reasonably powered Intel Core 2 Duo for what it needs to do.

The more annoying one is my desktop. I've had some oddities on boot recently, where it'd take a couple of restarts before it'd log in properly without hanging. I couldn't see anything obvious in any logs, but today it decided that my root partition needed checking. Then it decided that there were errors it didn't want to automatically handle and that I needed to fix it manually. Then it decided that there were lots of lost nodes, under-linked nodes (where the file table said it had more links than it could find) and overlinked nodes (where the file table said it had less links than it could find), multiply claimed nodes (where two files say they're different but refer to the same bit of disk), and so on. Before I managed to finish with that it said that libc was missing - which is the core of the C libraries, which are somewhat vital on most computers!

I've not got a clue what has happened to the disk and why it is failing, or why my other two partitions report clean, but I'm currently copying what I can in case the disk has had it completely. I was hoping to get some work done this weekend, but with a combination of this, parenting and a party for our next-door neighbour's daughter then it looks like I might not get as much done as I'd hoped.

Gah, why can't they make perfect disks that run forever and never corrupt?

[snowblizz]

Gah, why can't they make perfect disks that run forever and never corrupt?

Can't sell you new ones then!

Cheaper to replace bad disk than fabricate good ones.

Etc etc etc

I can come up with bunch of different ones just OTTOMH. Benefits of and education in economics.

Once upon a time you could fix windows computers, nowadays it is impossible. They've hidden everything away somewhere and unless the solution is "didn't attach power cord" the Windows help can't help you anymore.

Faulty ram chip? And it actually reported it? I've run with bad memory some times but the memorycheck, well the boot up one, has never found any errors. I still wonder what it is for.

[IBBoard]

Yeah, if you use the right tools then faulty memory is easy to find. You just need to use MemTest as a bootable CD rather than trusting to anything Microsoft make

Maybe perfect disks is a bit optimistic, but ones that at least go into read-only mode on failure rather than screwing up your data would be nice The annoying thing is that it was the disk I bought brand new when my last one failed two years ago that was erroring, not the refurb that I got under warranty as a replacement.

I've done a disk check etc using the great Parted Magic live disk (I also needed it to restore the Windows boot partition on the wife's computer when I took the dual-boot Linux disk out). Try finding something like that which runs as Windows! Anyway, SMART (the HDD self-monitoring) reports that the disk is fine, even on a long test. I've checked the filesystem and repaired the errors, then re-installed those packages whose files got modified (RPM packages - the installer file format - have a great little feature that lets you verify the files on disk against what was originally installed). Hopefully all is okay now and it'll work, since Seagate's tools sound like they're basically just a wrapped version of SMART and they won't take a return if it doesn't give an error code. Shame I lost the whole weekend.

[IBBoard]

Gah, more computer problems. I had some odd behaviour last night (errors about offsets not matching, random segfaults on the GUI, more and more files failing to validate against the installer package when I tested them), then I tried booting into Parted Magic and it failed twice, so I tried the MemTest on my machine.

Within 30s I had errors from memtest. I shut down, took one stick out and tested the other - three passes later and it all reported clean. I then shut down and tested the other stick - three passes later and it all reported clean. Odd, since one of those sticks should have been bad based on the first run. Put both sticks back and test again - three passes later and it all reported clean.

I'm checking the hard disk now, but hopefully it was just incorrectly seated memory that was causing my problems and the act of taking it out to check them individual has led to them being properly seated in the slots.

[snowblizz]

Fun times!

Out of curiosity... do you know of any long term problems with frying part of the "contact surface" of a memory module (one of the "pinns"). I once managed to install one dimm module incorrectly once so it somehow fried a connector where it wasn't correctly placed. I didn't immediately realise why it smelled charred components when starting t up...

[IBBoard]

It depends how the contact touches afterwards and how bad the frying is. If it is just surface singing and there's a good enough contact then it might still run without an issue. Downloading a Linux util disk (like Parted Magic) and running MemTest would tell you

My computer seemed fine after researching the memory, although I do have a pre-warning-warning from SMART (one of the HDD's spinup counts hasn't exceeded its threshold yet, but it is getting closer, so it is warning me to keep an eye on it). While I was running the tests then the drive also got rather warm, so I might end up cloning my machine back to the original disk that had filesystem corruption but no reported SMART errors.

[IBBoard]

Gah! It gets worse. The segfaults/errors/general flakiness returned after being okay for a couple of days after I fixed it by unplugging and replugging the memory (again), only this time I'm getting a continuous beep on boot. That apparently (according to the manual) means that my graphics card isn't seated correctly, but I'm getting a perfectly good picture from it. I'm wondering if the motherboard is bad, but I don't know how to make sure before I send it back. I've got a ticket in with Gigabyte, so hopefully they'll be able to shed some light on it :\

[IBBoard]

I found an easy solution - unplug the internal speaker! It's not normally a good idea to ignore an error, but in this case a) I want to use the computer, b) it still boots up fine, even though it is set to halt on anything except keyboard errors and c) if I unplug the graphics or the memory completely then I get a different set of beeps with a different tone, followed by the long beep. I've also tested with a different graphics card, so it shouldn't be graphics issues.

Hopefully I'll get a response from Gigabyte and be able to fix it, or I'll RMA it and get a replacement without too much of a wait.