Overhead compared to other methods

[eownerdead]

I took a (piece-of-crap) benchmarks of chroot, proot, fakechroot and bubblewrap.
It shows that fakeroot and proot are nearly twice as slow as chroot, but fakechroot is a bit faster.
I assumed that fakechroot uses LD_PRELOAD, so it has less overhead than proot, which uses ptrace, though.
What make fakechroot slow?

[josch]

Set the environment variable FAKECHROOT_DEBUG to 1 and you will see the overhead from translating paths for most library calls.