Upstream processed exec implementation

[ZacSweers]

This introduces a new ProcessedExecCli program that can be used to execute commands with a layer of added tracing. This CLI does the following

• Executes a given command
• If the command fails
  • Notifies an issue to Bugsnag
  • Checks if the command should be retried (possibly with a delay). If it can be, it's retried once.
• If the command is successful, it's just a passthrough layer
• Can read a config JSON file like so

   {
     "version": 1,
     "known_issues": [
       {
         "message": "Message shown in bugsnag",
         "log_message": "Message logged in CI, more detailed.",
         "matching_text": "The CI log text to match",
         "grouping_hash": "a-unique-string-for-bugsnag-grouping",
         "retry_signal": {
           "type": "delayed",
           "delay": 60000
         }
       }
     ]
   }

Usage

Use this in a local CI kts script and then pass it args

processed-exec.main.kts

#!/usr/bin/env kotlin
@file:DependsOn("com.slack.cli:kotlin-cli-util:x.y.z")

fun main(args: Array<String>) = ProcessedExecCli().main(args)

$ ./processed-exec.main.kts ./your-real-ci-script.sh

Future work

• Add issues that can fail a passing build (we've had cases in the past where the presence of certain logs are things we want to treat as an error).
• Configurability of how many retries.

[adamsp]

This output is wrong, if it exited with 0 we wouldn't be here. We also exit this process with 0 always (line 111), do we want to reflect the actual result of the command or do we always want to succeed?

[ZacSweers]

Good catch e8c3d17

[adamsp]

1 minute may not be correct, since we pull the delay from the retry signal

[ZacSweers]

0930612

[adamsp]

This is nice