Re: Github Actions

2019-08-28 Thread sebb
I think the pre-verified code could run on a separate system with restricted access. That's how self-service works for creating mailing lists, for example. In this case, there would need to be a separate host with read access to Jenkins. It could accept publish requests from Jenkins, and route the

Re: Github Actions

2019-08-28 Thread Francis Chuang
Hi Greg, Yes, the token is needed to push to calcite-site as we are running the job in the calcite repository. In terms of the token, if we are pushing to the calcite-site using Gitbox, then the token would be the one that is currently used in the git-websites node on Jekins. If we want to pus

Re: Github Actions

2019-08-28 Thread Matt Sicker
Secrets masking is another thing that only works for basic cases. Don’t try encoding it and printing it, for example. Just a different character set can throw off some tools let alone base 64. On Wed, Aug 28, 2019 at 04:23, Francis Chuang wrote: > Hi Greg, > > Yes, the token is needed to push to

Re: Github Actions

2019-08-28 Thread Joan Touzet
Continuing the top-post trend... I'd rather see full audit logs kept ~forever for any use of credentials, including the code that was executed. If we can't stop the leak, we can at least keep the paper trail. Right now, with our aggressive build cleanup steps, I don't think this is happening. Ar