The test_seed_random_data test specifies this user-data:
#cloud-config
random_seed:
data: 'MYUb34023nD:LFDK10913jk;dfnk:Df'
encoding: raw
file: /root/seed
and asserts that that is the exact content of /root/seed. If we're using a datasource which sets random_seed in its metadata (i.e. Azure and OpenStack, at least), this test fails: cc_seed_random will _append_ the cloud's random_seed to the given data before writing it out.
To be clear, the bug here is that this is unexpected: we do not document this behaviour. I can't think of a compelling case for not behaving in this way: more entropy is good, and we have write_files if users really do want to write out exact content to a path.
The test_seed_ random_ data test specifies this user-data:
#cloud-config LFDK10913jk; dfnk:Df'
random_seed:
data: 'MYUb34023nD:
encoding: raw
file: /root/seed
and asserts that that is the exact content of /root/seed. If we're using a datasource which sets random_seed in its metadata (i.e. Azure and OpenStack, at least), this test fails: cc_seed_random will _append_ the cloud's random_seed to the given data before writing it out.
To be clear, the bug here is that this is unexpected: we do not document this behaviour. I can't think of a compelling case for not behaving in this way: more entropy is good, and we have write_files if users really do want to write out exact content to a path.