| Commit message (Collapse) | Author | Age |
|---|
| |
|
|
| |
This entails moving a bunch of things into more appropriate modules, as well.
|
| | |
|
| |
|
|
|
|
| |
The explicit panic is a conscious decision; clap doesn't have a better way to represent mutually-exclusive options, so we have to use two boolean fields, but it will actually prevent the user from passing both options and triggering that code path. Since it's one of the rare situations where design choices converge on a truly unreachable case, I thought I'd make it easier to discover when I'm wrong and it is actually reachable.
The old logic treated (true, true) like (false, false).
|
| |
|
|
| |
This also lets me remove the clever-but-silly Single trait.
|
| | |
|
| |
|
|
| |
This comes with some substantial removals, because the AWS libs no longer treat every last field as optional when it's inappropriate to do so. Hooray!
|
| |
|
|
| |
This was an oversight; I missed that this trait existed!
|
| |
|
|
| |
sporadically the last one
|
| | |
|
| |
|
|
| |
an ASG.
|
| |
|
|
| |
It didn't add much to have our own Result definition, over using `anyhow::Result` directly.
|
| |
|
|
|
|
| |
This allows me to stop exporting symbols that were only exported for the purposes of making the CLI work.
This commit makes the minimum version for the next release 0.2.0, as it retires multiple symbols that are part of the lib crate's API in v0.1.x.
|
| |
|
|
| |
Using iterator combinators here feels like it's more in line with how I think about the problem.
|
| |
|
|
| |
The TTL also controls how long a service is unavailable because of cached DNS entries. For most services, a brief caching interval of 60s is more than adequate.
|
| |
|
|
|
|
| |
The original version of this program (written for another project) laid claim to the entire subtree under the target DNS name, as part of a broader feature of adding both per-instance hostnames and a per-ASG hostname with multiple records. This program doesn't do that; however, I forgot to remove this behaviour when porting the code.
This was caught in production, of course, though without any damage. An attempt to manage `grimoire.ca` failed because the AWS identity it ran under didn't have permission to do anything other than modify A and AAAA records.
|
| |
|
