Reduce amount of data downloaded from GitHub in resources/rebuild.sh

Use a tree-less git checkout [1], which reduces the amount of data
downloaded by ~60% (from >3GiB to ~1GiB currently) when cloneing the
upstream amazon linux repository from GitHub.

It will now download some stuff during checkout, but this is
significantly less than a full clone, as we're only checking out a
handful of tags.

We sadly cannot use a `--depth=1` checkout because we need access to all
tags, in chronological order, and for this, we need at least the commit
metadata itself fetched locally (even git ls-remote --tags will not sort
tags without first fetching these).

[1]: https://github.blog/open-source/git/get-up-to-speed-with-partial-clone-and-shallow-clone/

Signed-off-by: Patrick Roy <[email protected]>

Author: roypat

resources/rebuild.sh

@@ -120,7 +120,7 @@ EOF
 }
 
 function clone_amazon_linux_repo {
-    [ -d linux ] || git clone https://github.com/amazonlinux/linux linux
+    [ -d linux ] || git clone --no-checkout --filter=tree:0 https://github.com/amazonlinux/linux
 }
 
 # prints the git tag corresponding to the newest and best matching the provided kernel version $1
@@ -145,7 +145,8 @@ function build_al_kernel {
     local KERNEL_VERSION=$(echo $KERNEL_CFG | grep -Po "microvm-kernel-ci-$ARCH-\K(\d+\.\d+)")
 
     pushd linux
-    make distclean
+    # fails immediately after clone because nothing is checked out
+    make distclean || true
 
     git checkout $(get_tag $KERNEL_VERSION)