@ghostsquad We have had a related discussion of this with #747

I believe a concensus was reached whereby we would remove non-zero exit codes on commands that did not fail, but did not change the state of things (due to an existing plugin as you suggest). As noted in #747 (comment) the Issues and PRs that instituted this behaviour did not receive much feedback from our userbase (which was admittedly smaller) at the time.

@asdf-vm/core Are there any objections to addressing this issue with the proposition here #747 (comment)

EDIT: #66 also discusses this issue (I believe it was the first discussion)

If I'm reading the original issue (#322) correctly, it seems that @the-mikedavis also preferred the status code of asdf plugin add to be 0.

Pinging @asdf-vm/core / @Stratus3D again, would there be any objections to a PR that changes the exit code back to 0? As mentioned in the discussion that @jthegedus linked above there seem to be more requests for a successful exit code than an error.

Totally agree. At least give us an option (e.g. similar to --failure-exit-code in rspec)

I made a comment in the other issue, but it looks like it was closed. My code is littered with asdf plugin add <blah> || true and error handling suffers because of it, because if there's a problem getting the plugin, that error is essentially ignored.

This is a popular request and have linked a PR that resolves this - hopefully it lands in v0.12.1 soon.

For now, the best workaround is manually checking exit codes:

asdf_plugin_add() {
  asdf plugin add "$@" || {
    local exit_code=$1

    # If asdf detects that the plugin is already installed, it prints to standard error,
    # then exits with a code of 2. We manually check this case, and return 0 so this
    # this script remains idempotent, especially when `errexit` is set.
    if ((exit_code == 2)); then
      return 0
    else
      return $exit_code
    fi
  }
}

asdf_plugin_add <blah>

Theoretically, if Git returns with an exit code of 2 if cloning fails, or if a plugin exits with a code of 2, this will give a false positive, but the occurrence of said false positives is still drastically less than compared to without this workaround.