Dear Morello Technical Forum,
I followed the walkthrough video from Linaro (https://www.morello-project.org/resources/morello-platform-model-and-android-stack-walkthrough/) to install and bring up the nano Android stack on the Morello FVP.
Everything went as expected until I reached the Android stack build step where I received a FAILED message, as per the below snapshot:
"soong bootstrap failed with: exist status 1
ninja: build stopped: subcommand failed."
My host machine is a 2006 MacBook running Ubuntu 20.04.2 LTS.
Any help you could provide will be gratefully received.
Many thanks,
George
(Digital Security by Design Programme)
Hi George,The build process appears to be unexpectedly interrupted ("killed") by the operating system.This most commonly happens when there is not enough free memory available for the system to operate correctly. The operating system mechanism responsible for that is called "OOM killer" ("out-of-memory killer")
Could you please clarify the amount of RAM available in the system and amount of free disk space on the partition where the build is started?It should be possible to figure out these details with the help of the following command lines on an Ubuntu system:
# amount of RAM available free -h # amount of free disk space (when run from within the source tree directory) df -h .
When "OOM killer" interrupts a process, it would normally report relevant details at the system journal, which is available via:
journalctl -r
This information might be helpful to understand the failure.
There is a prototype of optimisation for building the nanodroid stack. The optimisation isn't published yet, however once it is published - it might potentially help with this via reducing the required amount of RAM and of free disk space.
Best regards,Ruben
Hi Ruben,
Thank you for coming back to me. That's very helpful.
On your raised points:
I have attached a relevant snapshot from the system journal for if it helps you to deduce any further info on the failure:
It's good to know you/Arm have an optimisation in the works for resource constraint machines. Would this come under Linaro's gitlab when it's made available?
Would I stand a better chance if I had a go at the build script on a machine with 8GB RAM available? Or it's subject to the overall configuration?
Kind regards,
Hi George,
The system journal output confirms that the build failure is caused by the "OOM killer".
While the 292GB of disk space should certainly be sufficient for the build, 3.8GB of RAM wouldn't be enough unfortunately.
> Would I stand a better chance if I had a go at the build script on a machine with 8GB RAM available?
The official minimum requirement at the moment is 16GB RAM.It certainly depends on the overall configuration and it might work with 8GB RAM, however it would be difficult to guarantee that in general case.
Kind regards,Ruben