Matt Martz comments

Results 336 comments of


                                            Matt Martz

copy module: modify time on the file (mtime) is not updating

I think I see what is going on. Due to moving to `shutil.copystat`, we're preserving the `atime` and `mtime` of the dest also, which we probably shouldn't. Can probably fix...

Can ansible produce friendly messages?

Duplicate of https://github.com/ansible/ansible/pull/79634

Plugin development documentation requires updates which address handling Unsafe value wrappers

Can you explain more about how you are using the data? If you are using the JSON and YAML encoders included with ansible, then it should be handled for you....

Plugin development documentation requires updates which address handling Unsafe value wrappers

```pycon >>> import yaml >>> from ansible.utils.unsafe_proxy import AnsibleUnsafeText >>> from ansible.parsing.yaml.dumper import AnsibleDumper >>> yaml.dump({'foo': AnsibleUnsafeText('bar')}, Dumper=AnsibleDumper) 'foo: bar\n' ```

Send _execute_meta results through the queue

Reproducer... ``` . ├── ansible.cfg │ │ [defaults] │ │ forks = 2 ├── hosts │ │ localhost0 │ │ localhost1 │ │ localhost2 │ │ localhost3 │ │ localhost4...

Ensure meta tasks don't disrupt host_pinned host/fork affinity

Reproducer... ``` . ├── ansible.cfg │ │ [defaults] │ │ forks = 2 ├── hosts │ │ localhost0 │ │ localhost1 │ │ localhost2 │ │ localhost3 │ │ localhost4...

Ensure meta tasks don't disrupt host_pinned host/fork affinity

This implementation has a shortcoming, in that meta only tasks won't break out of the inner loop. This can result in an extra non-meta task executing early.

Ensure meta tasks don't disrupt host_pinned host/fork affinity

The last commit may fix the previously noted shortcoming

When fatal: happens when using iteration, all results are lost even if ignore_errors is true

The problem here, is that the `Missing sudo password` error is coming from the connection plugin, and it does this by raising `AnsibleError`. In the `TaskExecutor`, we do not catch...

task timeout failure should indicate timeout for programatic use

Instead of inspecting the msg, we should just add a new `timeout=True` or similar to the response. The original ask was a new callback event, but we instead said we...