Strange dropping of credentials when using heroku cli

This build https://circleci.com/gh/Financial-Times/next-tme-search-api/180 fails to execute a step where it calls the heroku api using the heroku cli: https://github.com/Financial-Times/next-build-tools/blob/master/lib/enable-preboot.js. By the time it gets to this step of the build it has already successfully interacted with the heroku api several times, so authorisation is obviously ok. sshing into the circel VM to run the same task also works ok.
Not sure why this would be happening.

shellpromise: about to spawn heroku features:enable -a ft-next-tme-search-api preboot
shellpromise: output: Enter your Heroku credentials.

Fixed. Someone else had set up heroku deployment using the default circleci ui. This was resetting credentials to ones invalid for the actions I was carrying out with the cli.

Bit of feedback for the circle team - please make it possible to remove a deployment integration even if the person whose credentials are being used for it aren’t your own. If this team member had already left the organisation we’d have been left with a permanently broken build. Also maybe make the message ‘Install Heroku API credentials’ in the console output a bit clearer as to why it’s there.

Glad you got this to work, could you make this a separate Feature Request so that we can track it a bit better? :slight_smile: