We save off the top commit into a new file called 'gitrev'. This will help
us automate exploded tree creation by providing a simple file to read to
get the exact upstream commit to use as the base.
Before it would just error out because it was only looking at the patch-
listings in the sources file. So now we let it do that, and if the VER
variable is null after we just assume it's the first git snapshot of the
merge window and figure out the version from the linux- tarball.
Why I didn't fix this sooner, I have no idea. Probably because it only
happens 4 times a year and laziness wins the day.
There are two reasons for this. The first is that there is much more
context around the changes. The second is that this will make automating
an exploded git tree much easier to manage.
Upstream e1c972b681bf118fcedb9fe2ed7a73de983aa5ef makes it depend on
UIDGID_CONVERTED which is only set when all of the subsystems have been
converted to be user namespace safe. That defaults to Y whenever it happens,
so we'll set this after that point.
This cuts make prep time down from 53s to 10s on my desktop.
A full run across all archs can be done from within a source
tree by running ../../scripts/allarchconfig.sh