0.9.2: make fails on ARM w Makefile:827: recipe for target 'main.o' failed #4384

I checked out 0.9.2 today using github clone of 0.9.2 and tried to build.

pi@raspberrypi ~ $ uname -a
Linux raspberrypi 3.12.22+ [#691](/bitcoin-bitcoin/691/) PREEMPT Wed Jun 18 18:29:58 BST 2014 armv6l GNU/Linux
pi@raspberrypi ~ $ make -v
GNU Make 4.0
Built for arm-unknown-linux-gnueabihf

I made one change to the source, based on recco from #4362, but errors don't seem related to that.

$ ./autogen.sh; ./configure --disable-wallet --without-gui

Log from configure: http://pastebin.com/SdsGj4EG

$ make
$ tail -f bitcoin/makelog.txt
Making all in src
make[1]: Entering directory '/home/pi/bitcoin/src'
make  all-recursive
make[2]: Entering directory '/home/pi/bitcoin/src'
Making all in .
make[3]: Entering directory '/home/pi/bitcoin/src'
  CXX      addrman.o
  CXX      alert.o
  CXX      bloom.o
  CXX      checkpoints.o
  CXX      coins.o
  CXX      init.o
  CXX      keystore.o
  CXX      leveldbwrapper.o
  CXX      main.o
g++: internal compiler error: Killed (program cc1plus)
Please submit a full bug report,
with preprocessed source if appropriate.
See <file:///usr/share/doc/gcc-4.9/README.Bugs> for instructions.
Makefile:827: recipe for target 'main.o' failed
make[3]: *** [main.o] Error 4
make[3]: Leaving directory '/home/pi/bitcoin/src'
Makefile:849: recipe for target 'all-recursive' failed
make[2]: *** [all-recursive] Error 1
make[2]: Leaving directory '/home/pi/bitcoin/src'
Makefile:640: recipe for target 'all' failed
make[1]: *** [all] Error 2
make[1]: Leaving directory '/home/pi/bitcoin/src'
Makefile:497: recipe for target 'all-recursive' failed
make: *** [all-recursive] Error 1

I got make errors yesterday when I cloned "master" branch (yesterday was about missing files).

Are you sure you have enough RAM on you RPi to build bitcoind?

Not a bitcoin problem. As @sipa noted, it is probably RAM (check 'dmesg' for OOM killing evidence) or some other hardware issue. e.g. I've seen GCC ICEs like this when a CPU overheated.

@jgarzik - wow that was quick... 5 minutes to close a suspected issue based on "probable" OOM cause...

I've submitted a 4-5 bitcoin bugs that occured on this Pi with v0.9.1, so unless something spectacularly changed, it's not an out of memory condition as I've built bitcoind on the same box 3-4 times already. 512 MB RAM isn't little (okay, 16 GB is shared to display card).

Whatever the problem, it has nothing to do with bitcoin.

And yes, compiling one bitcoin source file may require many hundreds of megabytes of RAM. https://bitcointalk.org/index.php?topic=304389.0

Yes, but if I took time to properly submit this issue (and other already confirmed issues), that means I'm contributing. You could have added exactly the same comments and leave the issue open, then I'd come back and look into your comment, potentially confirm that you were right, you'd close the issue and others would be able to learn something.

So, the problem was the lack of virtual memory because two days ago I replaced the SD card and didn't increase swap size on the new card. By increasing swap size to the size I had before (500 MB) I was able to build bitcoin successfully as before.
A secondary conclusion of potential interest to bitcoin builders on small devices is that make can throw weird and unrelated errors - like file not found and recipe not found.

I'm adding these notes for the benefit of others only because you replied.

I can confirm that this is a out-of-memory issue. For this reason I added the memory recommendation to doc/build-unix.md at some point: 4babd08.

For small devices you should really use cross-compilation. You're tormenting yourself by compiling something as heavy as bitcoind on the device itself.

100% out of memory issue - Confirmed by checking dmesg (this happens on altcoins too): [255397.620182] Out of memory: Kill process 14776 (cc1plus) score 321 or sacrifice child [255397.620401] Killed process 14776 (cc1plus) total-vm:738196kB, anon-rss:473340kB, file-rss:0kB