[SOLVED] Old computer crashes while compiling certain packages with SlackBuild on Slackware 14

Neyzan · 10-23-2013, 03:29 PM

I have installed Slackware 14 with Fxce DE on my 11 year old desktop (1Gb RAM, Intel Celeron processor). It runs smoothly generally, but while compiling packages like xulrunner, chromium and wxPython with SlackBuild I get an unusually long string and at some point the computer crashes, without having finished the compilation.
Does anyone know why this is happening and how to solve it?

camorri · 10-23-2013, 04:07 PM

You have not provided any error information. Do you see any errors? Can you post them?

Just a guess. Open a terminal and run the command 'df'. Look to see where your temp file space is mounted. You have a gig of ram, and the default, if you are mounted on /dev/shm is half the available ram. Here is what I see on a 3 gig system.

Quote:

tmpfs 1.5G 116K 1.5G 1% /dev/shm

I have had a problem of running out of temp file space on larger compiles. I had to change fstab to mount a real hard drive dir, with lots of space to get around the problem. Just a guess remember, without any error info.

rkfb · 10-23-2013, 04:10 PM

I guess it would be difficult to say without seeing what actually happened or what the output was.

It would be a pretty lengthy process running any of these slackbuilds, more so on an older machine, they're not the simplest of builds any of them.

Are there not pre-compiled packages you could use? There's certainly xulrunner ones around I know that, someone's probably got the others somewhere.

volkerdi · 10-23-2013, 04:24 PM

Quote:

Originally Posted by Neyzan

I have installed Slackware 14 with Fxce DE on my 11 year old desktop (1Gb RAM, Intel Celeron processor). It runs smoothly generally, but while compiling packages like xulrunner, chromium and wxPython with SlackBuild I get an unusually long string and at some point the computer crashes, without having finished the compilation.
Does anyone know why this is happening and how to solve it?

Yes, your machine is running out of RAM in the linker stage. If you run htop in another console you'll be able to watch the free RAM drop to zero. If you're not using the SMP kernel with PAE support, switching to that (and having a lot of swap) might help. However, there is a process limit at play here, so it isn't as if having enough RAM/swap will necessarily fix the issue on x86.

Some of the Mozilla applications that bundle libxul are actually about to run into a wall on 32-bit machines where it might not be possible to compile them at all. There are better references out there than this one, but my Google-Fu is failing me right now:

http://mozilla.6506.n7.nabble.com/ld...-td280263.html

Neyzan · 10-24-2013, 11:14 AM

Sorry, but as my computer shuts down completely while compiling I can't provide any error information (unless there exists some way to retrieve it). I will try your suggestions though, and tell you if they work...

camorri · 10-24-2013, 12:58 PM

Quote:

as my computer shuts down completely

Do you mean it powers off? Can you explain a little more, exactly what happens.

Didier Spaier · 10-24-2013, 01:14 PM

Compiling huge programs can cause overheating and if the temperature is too high the machine simply halts to protect hardware. This happened to me a couple of times, but on a laptop with acpi enabled. I don't know if that can happen to a desktop

You could try to check the logs for that

EDIT. Example:

Code:

bash-4.2# grep  "Critical temperature" /var/log/syslog*
/var/log/syslog.3:Sep 22 22:07:02 T61 kernel: [   78.833940] Critical temperature reached (100 C), shutting down.
/var/log/syslog.3:Sep 22 22:07:02 T61 kernel: [   79.089174] Critical temperature reached (100 C), shutting down.
/var/log/syslog.3:Sep 22 22:07:11 T61 kernel: [   88.691847] Critical temperature reached (100 C), shutting down.
/var/log/syslog.3:Sep 25 22:44:04 T61 kernel: [20455.376716] Critical temperature reached (100 C), shutting down.
/var/log/syslog.3:Sep 28 11:43:52 T61 kernel: [   25.283926] Critical temperature reached (100 C), shutting down.
bash-4.2#

PS as those messages are from the kernel, maybe acpi is not needed if the temperature is acquired by sensors and transmitted to some driver. Does someone know how that works?

gnashley · 10-25-2013, 12:44 AM

You need a huge amount of RAM+swap in order to build chromium -I think they recommend something like 11GB.

mattca · 10-25-2013, 09:22 AM

Quote:

Originally Posted by gnashley

You need a huge amount of RAM+swap in order to build chromium -I think they recommend something like 11GB.

Can't speak to memory usage, but on a newer machine building the chromium slackbuild took several hours... I want to say at least 4. It's definitely a resource-intensive task.

Neyzan · 10-25-2013, 05:38 PM

What happens is that during the compilation of certain packages my computer just powers off. No warning, no error message, nothing. I checked the syslog but there is no mention whatsoever of the usage of my computer at the times that it powered off. I do get rather often the message

Code:

udisksd[some 4-digit number]: Error performing housekeeping for drive /org/freedesktop/UDisks2/drives/WDC_WD800JB_00ETA0_WD_WCAHL3464188: Error updating SMART data: sk_disk_smart_status: Input/output error (udisks-error-quark, 0)

but I don't know if it is relevant.
I have a swap space of 2Gb now. I will change a 15Gb partition I'm not using into swap and see if this will solve the problem...

camorri · 10-25-2013, 06:12 PM

If the machine is powering off, it is probably a hardware issue. I have never seen a failed compile cause a machine to power off.

Things to check, are all the fans running at full speed? Are there blocked vents on the system? Dust build up? Since it is a desktop, remove the covers and have a good look around. Most systems will run without covers installed. You can visually check fans and dust.

ReaperX7 · 10-25-2013, 10:47 PM

Xulrunner can use, as Patrick said, up to about 6GB of RAM/Swap space in the linker stage. Often during compiles the CPU is under a stress load also which could make your temperatures shoot up also.

You might want to check your fan unit and see if it's dusty or dirty and take a can of compressed air and clean it and the case out. Also, get some Arctic Silver thermal grease and put some on your CPU's block chip and reattach the fan unit.

If it continues and you may need to replace the fan unit, or allow more ventilation (open the case or add extra fans if possible).

AlienBOB has pre-built packages up on his server and SalixOS has pre-built packages also so you shouldn't need to build packages unless necessary.

yars · 11-02-2013, 03:41 PM

Possibly I is not right, but I think you should adjust the SlackBuild script in line that contain the MUMJOBS variable. By default that variable contain a big number of threads, like 7. Old processors does not have these number of threads and when you set it in high, building process can be terminated with the out of memory errors.

TobiSGD · 11-02-2013, 09:02 PM

This indeed sounds like your machine is overheating. Clean out the system, exchange the thermal paste on the CPU, check the temperatures.

andrewthomas · 11-02-2013, 10:10 PM

Quote:

Originally Posted by mattca

Can't speak to memory usage, but on a newer machine building the chromium slackbuild took several hours... I want to say at least 4. It's definitely a resource-intensive task.

Code:

andrew@msi-gentoo ~ $ qlop -t chromium
chromium: 3594 seconds average for 20 merges
andrew@msi-gentoo ~ $ qlop -t xulrunner
xulrunner: 1302 seconds average for 12 merges

This is on a first gen Phenom with 4G RAM.

Code:

vendor_id	: AuthenticAMD
cpu family	: 16
model		: 2
model name	: AMD Phenom(tm) 9500 Quad-Core Processor