Release-Script: Slow, single-threaded xz compression of factory images

[kuleszdl]

The last line of the release script does only use one cpu thread for compressing the factory image:

xz -v --lzma2=dict=512MiB,lc=3,lp=0,pb=2,mode=normal,nice=64,mf=bt4,depth=0 $DEVICE-factory-$BUILD_NUMBER.tar

In theory, xz supports multiple threads using the -T parameter. I tried setting it to -T0 but then xz failed as it was out of memory (I tried in a VM with 30G of RAM and 15 cpu cores). Is there a way how this could be improved?

[thestinger]

It's intentionally using a huge dictionary and spending a lot of time compressing because it only happens for release builds, not development builds. I have more than enough memory to use some threads but that would hurt the compression ratio.

You can just comment it out. In fact, you don't need to generate factory images at all after your initial build so you can comment out everything but generating over-the-air updates.

[thestinger]

It's intentionally using a huge dictionary and spending a lot of time compressing because it only happens for release builds, not development builds. I have more than enough memory to use some threads but that would hurt the compression ratio.

You can just comment it out. In fact, you don't need to generate factory images at all after your initial build so you can comment out everything but generating over-the-air updates.

[thestinger]

And technically, for the first build, you didn't need to generate an over-the-air update.

[thestinger]

And technically, for the first build, you didn't need to generate an over-the-air update.

[Rudd-O]

Glad to take patches here https://github.com/Rudd-O/copperheados-build/ to skip certain steps via build variables.

[Rudd-O]

Glad to take patches here https://github.com/Rudd-O/copperheados-build/ to skip certain steps via build variables.