maemo.org - Talk - Compiling on-device problem

maemo.org - Talk (https://talk.maemo.org/index.php)

- SailfishOS (https://talk.maemo.org/forumdisplay.php?f=52)

- - Compiling on-device problem (https://talk.maemo.org/showthread.php?t=95723)

Compiling on-device problem

After managing to compile QtCreator on-device everything seemed to be well and good and awesome, but today got this error:

Code:

g++ -c -pipe -O2 -Wall -W -D_REENTRANT -fPIE -DQT_NO_DEBUG -DQT_QUICK_LIB -DQT_WIDGETS_LIB -DQT_GUI_LIB -DQT_QML_LIB -DQT_NETWORK_LIB -D

QT_SQL_LIB -DQT_CORE_LIB -I/usr/share/qt5/mkspecs/linux-g++ -I. -I/usr/include/qt5 -I/usr/include/qt5/QtQuick -I/usr/include/qt5/QtWidge

ts -I/usr/include/qt5/QtGui -I/usr/include/qt5/QtQml -I/usr/include/qt5/QtNetwork -I/usr/include/qt5/QtSql -I/usr/include/qt5/QtCore -I.

-o qrc_resources.o qrc_resources.cpp

g++: internal compiler error: Killed (program cc1plus)

Please submit a full bug report,

with preprocessed source if appropriate.

See <http://bugs.merproject.org/> for instructions.

make[1]: *** [qrc_resources.o] Error 4

make[1]: Leaving directory `/home/nemo/dev/cool-retro-term-master/app'

make: *** [sub-app-make_first-ordered] Error 2

The source of the prog is here:
https://github.com/Swordfish90/cool-retro-term
(you also need the qmltermwidget)

Is that jolla killing process too soon (on first try had it in background and was doing something else, second attempt foreground nothing running except term, same result), or gcc bug? Bit worried, had programs with 12+ hours of compile time on n900 and it looks like the jolla watchdog is killing it :/

Re: Compiling on-device problem

ok, I think by now tried all the echo tricks from OOM thread, cc1plus still gets killed, not sure if should try swappiness tricks or just gcc bug :(

Re: Compiling on-device problem

Is dmesg showing something, e.g.

Code:

[88886.470155] lowmemorykiller: Killing 'genautomata' (10246), adj 0,

   to free 503056kB on behalf of 'kswapd0' (50) because

   cache 16164kB is below limit 16384kB for oom_score_adj 0

   Free memory is 11948kB above reserved

-> All memory+swap exhausted.

Re: Compiling on-device problem

Yeah, well damn, it's like with scummvm that worked in 1.2.0.5 and stopped with 1.3.0.8, one step forward two steps back (my bet qtcreator is no longer compilable on device):

Code:

[nemo@Jolla ~]$ dmesg | grep lowmem

[ 5942.342712] lowmemorykiller: Killing 'harbour-radio-s' (6416), adj 71,

[ 5942.752326] lowmemorykiller: Killing 'fingerterm' (1289), adj 70,

[ 5944.308224] lowmemorykiller: Killing 'QQuickXmlQueryE' (14427), adj 70,

[ 5951.088172] lowmemorykiller: Killing 'QQuickXmlQueryE' (14427), adj 70,

[ 5952.194627] lowmemorykiller: Killing 'QQuickXmlQueryE' (14427), adj 70,

[ 5953.207964] lowmemorykiller: Killing 'QQuickXmlQueryE' (14427), adj 70,

[ 5954.442941] lowmemorykiller: Killing 'QQuickXmlQueryE' (14427), adj 70,

[ 5955.485883] lowmemorykiller: Killing 'QQuickXmlQueryE' (14427), adj 70,

[ 5957.336699] lowmemorykiller: Killing 'QQuickXmlQueryE' (14427), adj 70,

[ 5958.812634] lowmemorykiller: Killing 'QQuickXmlQueryE' (14427), adj 70,

[ 5960.560872] lowmemorykiller: Killing 'QQuickXmlQueryE' (14427), adj 70,

[ 5961.918814] lowmemorykiller: Killing 'QQuickXmlQueryE' (14427), adj 70,

[ 5963.222187] lowmemorykiller: Killing 'QQuickXmlQueryE' (14427), adj 70,

[ 5964.705569] lowmemorykiller: Killing 'QQuickXmlQueryE' (14427), adj 70,

[ 5965.718479] lowmemorykiller: Killing 'QQuickXmlQueryE' (14427), adj 70,

[ 5966.662321] lowmemorykiller: Killing 'cc1plus' (15149), adj 0,

[nemo@Jolla ~]$

so even with:
echo "0,0,0,0,0" > /sys/module/lowmemorykiller/parameters/minfree
cc1plus gets killed *sigh*

Re: Compiling on-device problem

Quote:

Originally Posted by nieldk (Post 1476566)

Look here.
https://answers.launchpad.net/ubuntu/+question/236764

with

Code:

echo '9999' > /sys/module/lowmemorykiller/parameters/adj

echo '1' > /sys/module/lowmemorykiller/parameters/minfree

Still cc1plus was killed :/
(interestingly enough, browser running in the background was not, weird, but compilation failed, previously pretty much everything was killed too and only fingerterm left)

Re: Compiling on-device problem

How about adding more swap?

Code:

devel-su

fallocate -l 2G /swapfile #2Gigs

chmod 600 /swapfile

mkswap /swapfile

losetup --find --show /swapfile 

# ->  /dev/loop0

swapon /dev/loop0

Re: Compiling on-device problem

Quote:

Originally Posted by AapoRantalainen (Post 1476583)

How about adding more swap?

Code:

devel-su

fallocate -l 2G /swapfile #2Gigs

chmod 600 /swapfile

mkswap /swapfile

losetup --find --show /swapfile 

# ->  /dev/loop0

swapon /dev/loop0

perhaps. Or, might be tmpfs that is too small (I run in to this on a regular basis on N9). ?

Can be increased 'on the fly'
(40M in this example)

# mount -o remount,size=40M tmpfs /tmp/

Re: Compiling on-device problem

tmp stays at 400mb free throughout (was same on n900, only time needed to resize its tmp was for wifite which used to dump ivs in there)

will check out later with increased swap, hopefully this will do it (but such a shame lowmemkill isn't just toggleable, the too deep sleep at least can be turned off in one command):

Re: Compiling on-device problem

Ok, after watching in htop, looks like nieldk's answer was correct, the killing of background apps is inhibited and it is just gcc eating through whole swap and then remaining ram until it dies, going to check with increased swap, thanks!