Thanks for this, Alan. Most of the size from the Snapcraft cache is stage packages. We really wanted to bind-mount it, but the user mapping in lxd got in the way there so it wouldn't be usable. As a result we ended up copying it, not considering remote cleanbuilds where that would take quite some time.
In your mind, what is the ideal solution in this scenario? A better progress indicator as you suggest? Or no cache copying to remote hosts?
Thanks for this, Alan. Most of the size from the Snapcraft cache is stage packages. We really wanted to bind-mount it, but the user mapping in lxd got in the way there so it wouldn't be usable. As a result we ended up copying it, not considering remote cleanbuilds where that would take quite some time.
In your mind, what is the ideal solution in this scenario? A better progress indicator as you suggest? Or no cache copying to remote hosts?