test: Add tests for parsing logs from init commands, tests for parsing of RFC3339 timestamps with different levels of precision #180
Conversation
|
I might add an additional test that generates timestamps directly and shows no difference between the two formats in this PR hashicorp/terraform#37818 |
…ted timestamps from logs
SarahFrench
changed the title
github.com/hashicorp/go-hclog dependency to support this
radeksimko
left a comment
radeksimko
left a comment
There was a problem hiding this comment.
This is very reasonable assuming it stays test-time dependency only - which I'm assuming is the intention. Correct me if I'm wrong.
Avoiding the runtime dependency on this library is important because tfjson has so far been a very light-weight library that maintained minimal impact on consumers.
Increasing the runtime dependency surface increases the chances of all kind of unnecessary noise in security scanners and maintenance efforts everywhere and ultimately the library should not need it at runtime because it produces logs but it's not meant to consume them.
Yeah, good point. I'll replace the use of the dependency with some string vars that show what the time looks like formatted in both ways, and a code comment to make it clear what's happening. |
SarahFrench
dismissed stale reviews from radeksimko and mildwonkey
via
bac6efc
SarahFrench
changed the title
github.com/hashicorp/go-hclog dependency to support this
3 participants
Related Issue
N/A
Description
In hashicorp/terraform#34886
initcommands were given some support for JSON output using the-jsonflag. Log messages from the init command are a mixture of types"log"and"init_output".This PR adds some tests showing the current behaviour of how JSON logs from init are parsed. Currently terraform-json doesn't recognise the
init_outputtype of log, so those logs are parsed as an unknown message type.Also, this PR's test was made using output generated from Terraform using the code from PR hashicorp/terraform#37818. This demonstrates that parsing JSON logs is unaffected by that change, but just to be sure I added a test that generates timestamps in both formats and shows that both are parsed fine (though there are different levels of precision, expectedly).
Rollback Plan
Changes to Security Controls
N/A