The sandbox overrides HOME so git cannot find the user's global config,
causing 'Author identity unknown' errors when acceptance test scripts
make commits. Write a minimal .gitconfig with user.name and user.email
into the sandbox working directory during sharedSetup.
Co-Authored-By: Copilot <223556219+Copilot@users.noreply.github.com>
There is still a bit of work to get the gh pr comment tests in order, however this goes a way towards fixing the issue along with acceptance tests.
Also, it turns out some of the issue acceptance tests were really running `pr` tests.
While swarming on this, I've encountered many merge conflicts based on
where folks have introduced new test functions. Alphabetizing them should
reduce the number of merge conflicts we encounter while introducing more
of these tests moving forward
To test this, I decided to look into the .gitconfig used for the test and
examine the credential helpers. However, the format of the git command is
`git config --get credential.<URL>.helper`
What's awkward about this is that the <URL> depends on the host the user
specified when running the tests, meaning I'd need to create a key like
credential.https://github.com.helper to access what I need while setting
this up.
There was no functionality for string formatting before, so I added the
command formattedStringToEnv which essentially wraps fmt.Sprintf() and
saves the string to an environment variable.
This allowed me to dynamically create the config key in the test.
The previous commit introduced two new functions, setEnvVar and
deleteEnvVar that are duplicative of functionality native to testscripts.
This commit switches to the native functionality and removes the
duplicative functions introduced in the previous commit.
Additionally, it removes the `--token` flag that was added to
`gh auth login`
Additionally, this commit adds acceptance testing for `gh auth login` and
`gh auth logout`.
The --token flag was necessary for adding testing for `gh auth login`
because the current implementation with `--with-token` appears to be
broken. It hangs, waiting for user input, but user input doesn't exit it.
Additionally, it appears that `--with-token` is intended to allow for TTY
input of an auth token, but it isn't implemented.
`--with-token` does work when used with the redirect operator `<` when the
token is saved in a file. However, due to limitations of testscripts, I
could not use a file for saving the token in a repeatable manner. Thus,
implementing the `--token` flag seemed like a quick solution to validate
that the direction I was going during testing was valid.
Whether the flag stays or not is up for discussion, and I'd love to get
input on that from the team.
This is a bit of a refactor based on the work done in `gh workflow` as a better approach to verify secrets created are what we expect.
Changes made:
1. Removed `env2lower` as it wasn't being used in testscripts
2. Added `replace` custom command to deal with testing organization workflow secrets
3. Refactored secret testscripts to create and run workflow that tests the value of the secret provided
4. Minor reordering of test `acceptance` test functions as appending to the end is confusing and adds conflicts
5. Removed stdout TTY assertions
This commit refactors 2 new custom commands to work more like the native env command, allowing for multiple environment variables to be set.
Additionally, the documentation on acceptance tests have been expanded to list out custom commands.
These are initial testscripts for confidence around core `gh secret` subcommands around 1) repository, 2) repository environment, 3) organization secrets.
Each testscript exercises the 3 core subcommands of `gh secret`: 1) setting, 2) listing, and 3) deleting each type of secret.
Note should be made that repository environments do not exist for free accounts such as `gh-acceptance-testing`.
