Problems building LibreOffice on RPI3 Slackware hard float
 

I have tried building from source with all the listed dependencies from SBo built and installed and also tried using AlienBob's package but my RPI locks up after about 2-3 hours of chugging through.

Once it locks up it is frozen solid, can't even ssh into it remotely.

Because of this I have to just cycle the power so I can't give more details on exactly where it locks up.

I have the -current branch installed but I don't have KDE, is the lack of KDE perhaps causing the problems?

Thanks in advance for any suggestions.

Hi slackist,

I have done a lot of compiling with the RPi3 and from my own experience I can tell you that during long periods of hard work the SoC heats up to around 90 degrees Celsius and then exactly what you've described usually happens. On my RPi3 I have a "thermo alloy" heat-sink, 28mm x 28mm x 20mm approx. mounted on the top of the Soc with a M3 thermal pad to secure it, and I find that even that is not sufficient to keep it cool enough on occasion. I am planning to fix up an active (air) cooler which will just be a heat-sink with a fan bolted on top of it.

One other time I had an issue with a power supply on the RPi3 which produced similar lock-ups. The PSU was giving out exactly 4.8 volts and I swapped it for a PSU that gave out 5.1 volts @ 2A which solved the problem completely.

Have a check every few minutes to see how hot your SoC is getting. If it gets into the 80's Celsius then it won't be long before your RPi3 starts giving you problems, or just locks up and you have to reboot. If it's not the heat that's causing this problem then perhaps try another PSU.

Hope this helps.

Thanks for that.

I've run a few more SBo scripts but nothing quite as big as LibreOffice with a watch running on the temperature and it gets to around 60-61C but never higher than that.

I don't think I have any way to accurately check the power supply output voltage but I'm using a Raspberry branded one rated at 5.1V and 2.5A output for RPi3 however since I am in the land of fakes that might not mean much! I've had 3 microSD cards die already but I think I've managed to get a real one this time as it's booting and rebooting fine and not throwing up errors.

I have a neighbour with a digital multimeter so I'll see if I can borrow it and check.




PS: I just noticed your location is listed as Middlesbrough, small world indeed.. I grew up in Harrogate, may Bob have mercy on my soul.

Only 60 degrees? Wow. I wish mine stayed that cool. When I'm compiling with Slackware ARM on the RPi3 I use all four cores and hammer the hell out of mine for hours on end. It heats up very quickly. My RPi3 can halt/crash/freeze anytime, or not at all. It seems random but I've read somewhere that once the CPU hits 92 degrees Celsius it's 'game-over' and the system will halt to protect itself from serious damage. Seems logical I suppose.

Yes, location Middlesbrough... for my sins. From Halifax originally. Long story. Don't ask!

OK, I tried again.

Build from source with all the listed SBo dependencies either installed or disabled:
I managed to get it up to 67°C! But then it locked up at
[PY ] CustomTarget/writerfilter/source/ooxml/OOXMLFactory_dml-baseTypes.hxx

I am assuming it is a PEBKAC problem and I am missing something obvious because I am thick but has anybody got LibreOffice installed on a Slackware RPi3?

I wrote my own simple bash script, that's by no means accurate, which gives an indication of the temperature of the RPIs CPU. When I'm compiling I'm always running in a screen. So, if I have issues with crashes I'll start the compile, exit the screen, and leave it running, then I can run the bash code periodically to keep an eye on the temperature. Obviously you can mess about with it, write it to a filename.log, crontab it, do whatever suits your needs.

Heh, your bash script looks like the source code for HHGTTG compared to my inelegant and clumsy effort!
I ssh'd into the RPi from another box and ran watch on the script so I could see what it got up to before it locked up and 67 was the last reading before it went into paperweight mode.

Whatever works is good in my book. :D

Update on cooling -
This morning I mounted a 28mm x 28mm x 14mm heatsink, with a 30mm x 30mm x 10mm (5v) fan screwed onto it, on top of the SoC on my RPi3. Then I went ahead and did some hardcore compiling using all four cores for approx. 2-3 hours. I'm still on with it now and expecting it to complete shortly. The CPU temperature hasn't yet risen above 50 degrees Celsius and there's no indications that it will.

So, this heatsink and fan I bought from eBay very cheap is solving all my problems with overheating. You can tell how cheap it is when you hear the fan running. It sounds like a hammer-drill. I think the bearings in the fan are shafted. For now though, it's working. ;)

<edit> So, in the end the highest CPU temperature reading was:

This is a good 30 degrees Celsius lower than usual. Previously, by the time compiling is done the SoC will usually be well into the 80s Celsius. I didn't think air cooling would make this much difference, considering the size of the x86 chipset passive heatsinks I have been using. Now, it's generally said that the onboard temperature reading is always a few degrees higher than it actually is. That doesn't matter because it's the only datum I'm using or have used. So, these comparisons are good. There were no freezes, crashes, or other 'overheating' related issues throughout - which there usually are! Active cooling on the RPi3, for my situation, was the answer.

<another edit> If you're interested here's the links to what I've used.
Link to RPi3 heatsink/fan on eBay. This is the cooler I'm using. I connected the power wires to an old CD-ROM audio cable plug. I removed the pins inside and soldered them to the fan cables and re-inserted them back into the plug. Works great! I secured the heatsink to the SoC using 3M 14m x 14mm adhesive thermal pads from eBay, which fit perfectly. Oh, and I cut a 32mm x 32mm square piece of plastic from the see-through front of a donut box (I'm not kidding) and then cut a 16mm x 16mm square piece out of the middle and used that as an insulator between the RPI3 circuit-board and heatsink. I tested the plastic by dropping it in water I'd just boiled and it survived admirably. Sorry, but I don't have a link to the company who make those donut boxes.

You might also consider slowing down the CPU clock to, say, 900 or even 800 MHz. The build will run more slowly, but you won't lose time struggling with thermal issues. As root:

# cpufreq-set -u 900MHz

If you have an older RPi sitting around, you can wire the serial consoles together and use the older RPi to capture the console output.

# tee /dev/ttyAMA0 rpi3-overheat.log

One last suggestion: swap space might have developed a bad sector, which will crash any system, not just an RPi. (My desktop would lock up during kernel builds. I thought it was thermal, but it wasn't.) Re-formatting swap space(s) using "mkswap -c /dev/name" every so often might be a good idea.

Thanks gus3, but under-clocking the RPI3 is not going to happen. If I wanted it to run slower I'd use a RPi2. The RPi3 runs Slackware ARM fine when I'm just using it for testing or in Xfce. Only while compiling and using all four cores was it crashing. The problems I had were all to do with the SoC getting too hot and I've found an acceptable solution. After doing three sarpi3 test builds today, and currently working through the fourth, without so much as a hiccup, I think it's safe to say that my overheating issues are a thing of the past. The RPi3 compiles a lot quicker when it's cooler! I'd guesstimate at around 10-15% shorter compile times when the SoC is not getting too hot.

Re-formatting swap may be a point to consider but it's not relevant in my case. I've always known what the problem was and have tested various passive cooling methods to try and resolve it. Now I've found the time to put this active cooling shizzle together, it's not a problem anymore. ;)

OK, based on all the sound advice I did the only sensible thing and went online shopping! :)

Turns out that the heatsink will probably not fit under my rtc module even with the GPIO longlegs or whatever it is called fitted:

http://thaieasyelec.com/en/developme...pi-detail.html

Naturally after all this I had to add a breadboard, a few wiry things, one of those T-cobbler things and some other bits and pieces so that is my story and I am sticking to it (in case my wife asks).

Thanks to gus3 for the suggestions, I will try the formatting of swap also - that is an interesting idea.

When I first started testing with passive heatsinks on the RPi3 I hit the same problem. I was using a RTC similar to the one you've linked, but it was a ds3231 like this and not a ds1307. So, I was quite fortunate to have bought a ChronoDot RTC back in 2013 for the RPi1, which requires jumper cables like these to connect it to the GPIO instead of directly plugging in. I also have some of these ds3231 RTCs which connect via the same method. They are very easy to connect/disconnect with the jumper cables and you don't have a problem with the heatsink being in the way of the RTC. This is how I connect my RTCs to the Raspberry Pi 3 now since I have had progressively larger heatsinks getting closer and closer to the GPIO pins.

If your wife asks I will deny everything. ;)

This is driving me fsck nuts now.

I can build and install stuff from SBo (medit, sylpheed and others) but LibreOffice is a no-no.

I have tried (I think) everything that the interwebs suggest but I can't get it.

I solved the heat problem, a heatsink and a small fan blowing onto it shows me that the system doesn't get over 59-62C even under pretty heavy loads.

Does anybody have a pkg for LOffice that will work on RPi3 -current or any tips or suggestions?

Download the LibreOffice package from Salix: 100% Slackware compatible!

Will it install/build on ARM?