GCC intermittantly failing

passive · Tux's lil' helper Joined: 31 Dec 2004 Posts: 105

Hi All,

I have a system that's had a variety of errors over the past few months. Very inconsistent, generally, but there has always seemed to be something a little "off".
I just did a fresh install of Gentoo, and while updating the core packages, I kept getting errors. I was unable to rebuild GCC or glibc, for example, but there were also failures in a variety of other packages. Trying to install the most basic KDE, which is about 100 packages, including the ones for X, resulted in about a dozen of them failing.

Some of the errors specifically mentioned GCC failing, and cautioned that it was probably a hardware or OS problem. This would make sense given the history I've seen, but I'm not exactly sure how to track it down. I don't think it's a hard drive problem, as of the two in there, one is brand new, and the other has been fairly thoroughly tested. I suspected memory, and am currently running memtest+, but it's got 18 passes with 0 errors so far.

Are there any tools you could recommend to diagnose what might be failing? Thanks.

eyoung100 · Veteran Joined: 23 Jan 2004 Posts: 1428

Try using an emerge wrapper first to correctly rebuild your toolchain. I use emwrap. For more information see this post. If there is a bunch of cruft(file trash), or you've had your system up awhile, the toolchain sometimes can become unstable. If emwrap fails to rebuild the toolchain, then you can safely rule out gcc as being the cause. If this doesn't work, please reply with the manufacturer of your hard drives.
_________________
The Birth and Growth of Science is the Death and Atrophy of Art -- Unknown
Registerd Linux User #363735
Adopt a Post | Strip Comments| Emerge Wrapper

tarpman · Posted: Tue Oct 30, 2007 9:01 pm Post subject:

Overheating is a major possibility.
_________________
Saving the world, one kilobyte at a time.

eyoung100 · Veteran Joined: 23 Jan 2004 Posts: 1428

passive · Tux's lil' helper Joined: 31 Dec 2004 Posts: 105

Ok, well here's where I'm at:

Memtest+ went to 30 runs with 0 errors. Given how frequently I get errors when it's running regularly, I figured it's probably not the memory.

I tried:

emwrap.sh -t world

Which started with glibc, and segfaulted. I rebooted twice, an ran

emwrap.sh -r -t world

each time, with a segfault as my reward.

Now I'm looking into diagnosing the CPU. It's an Athlon 64 at 2.2 Ghz, and I have a very new fan on it. I'm currently reading this guide:

http://www.gentoo.org/doc/en/articles/hardware-stability-p1.xml

redgsturbo · Apprentice Joined: 24 Jun 2005 Posts: 283

eyoung100 · Veteran Joined: 23 Jan 2004 Posts: 1428

passive · Tux's lil' helper Joined: 31 Dec 2004 Posts: 105

I got lm_sensors working, or so it seems. Running CPUburn takes the k8temp sensor from 48 to 76 degrees in just a few minutes. I'm guessing this is the problem.

I don't think the CPU is overclocked, but I got the system as a whole from my brother, so it is a possibility. I'm not entirely sure of the thermal paste used, so I will double check that as well.

Thanks for all the help so far.

eyoung100 · Veteran Joined: 23 Jan 2004 Posts: 1428

Would you please paste your /etc/make.conf?? Let's make sure your compile settings are not overdone also.
_________________
The Birth and Growth of Science is the Death and Atrophy of Art -- Unknown
Registerd Linux User #363735
Adopt a Post | Strip Comments| Emerge Wrapper

passive · Tux's lil' helper Joined: 31 Dec 2004 Posts: 105

Ok, the CPU was overclocked. It's a 2800+ which is supposed to be at 1.8Ghz. Unfortunately, the only keyboard I have is a on imac USB model, without a delete key, so thus far I am locked out of the BIOS. Fortunately, I've reset enough BIOS' in my time to be able to guess the correct jumper, and I'm now back operating at 1.8Ghz.

I also cleaned off and reapplied the thermal paste, which I believe is Artic Silver Ceramic.

At this point, it idles around 44 degrees. I have a thermaltake Silent Boost fan, which probably contributes to this relatively high idle temp. Running CPUburn, it's still getting remarkably hot (I've seen 82 degrees so far, though it's taken far longer then before to get to that point), so I wonder if my fan isn't working properly. I did a bit of research before purchasing this model, and it should have no problem cooling this CPU. The heatsink itself gets very hot, so I think it is absorbing heat very well.

Oh, here's my make.conf (I've barely done anything with this system yet):

redgsturbo · Apprentice Joined: 24 Jun 2005 Posts: 283

eyoung100 · Veteran Joined: 23 Jan 2004 Posts: 1428

Below 100 F is perfectly acceptable. Try:
Change -march to k8 then:

passive · Tux's lil' helper Joined: 31 Dec 2004 Posts: 105

Oh, all those temperatures are Celsius.

At the moment, I think everything is ok. I'm compiling GCC, and the temperature is stable around 57 degrees. I will change the --march, and give that a try.

Thanks for all your help.

NeddySeagoon · Posted: Wed Oct 31, 2007 9:16 pm Post subject:

passive,

80C is a top limit - hotter and you can expect problems but its not a hard limit.

The hot heatsink indicates that the heatsink is in good thermal contact with the CPU but that the heat is not being conducted to the air, or the hot air in the case is unable to get out. The PSU fan and the rear case fan should both be moving hot air out of the case, so air flows in at the front bottom, diagonally across the motherboard, over the CPU and out at the rear.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

tarpman · Posted: Wed Oct 31, 2007 9:23 pm Post subject:

To take NeddySeagoon's idea one step further, check that vents at the front and back are clear of dust (and the CPU fan while you're at it).
_________________
Saving the world, one kilobyte at a time.

eyoung100 · Veteran Joined: 23 Jan 2004 Posts: 1428

Congrats on getting the compile started. Sorry for the confusion with the temperatures. I have a feeling the US will never be metric, oh well. Now that the temperature stays constant while working you're where you need to be. Each system that you build will have its own unique characteristics, one of which is temperature. The reason you could not trust your gut on what the temperature should be is because you received this from your brother. If the temperature stays constant at around 57C(135 F), it should never rise or fall more than 5 to 10 degrees in either direction.
_________________
The Birth and Growth of Science is the Death and Atrophy of Art -- Unknown
Registerd Linux User #363735
Adopt a Post | Strip Comments| Emerge Wrapper

passive · Tux's lil' helper Joined: 31 Dec 2004 Posts: 105

Ok, that worked out pretty well. I think the CPU has simply decided not to run at that speed anymore. It's fine with me, it's a server box.

At this point, I've had many packages fail compiling, is there an easy way to recompile everything?

Thanks again.

eyoung100 · Veteran Joined: 23 Jan 2004 Posts: 1428

type