Conversation
Signed-off-by: Chris Hupman <chupman@us.ibm.com>
01eb732 to
8fc3ebd
Compare
|
I was looking to add code coverage to travis when I spotting hard-coded credentials in the travis config. Since it's only 2 commits I think it's easier to just do 2 reverts and reapply once we scrub the token. I also don't think we want CI pushing out pypi releases, but that might just be me. |
|
@chupman : Ideally we would not want CI to push out into PyPi each time we push into a particular branch, or into The downside is that, there will be multiple overwrites of library corresponding to same version num in pypi. I don't know if that is clean way. Or other thing we can do is, we can completely remove the configuration in If we plan to continue with second option, the problem is that each time a new version is out, the task of publishing library to PyPi becomes a manual process, and someone has to do that manually. Let me know your thoughts? |
FlorianHockmann
left a comment
There was a problem hiding this comment.
Any reason why we need a dedicated 0.1.0 branch instead of just publishing from master?
Regarding deployment from Travis: It can be set up to only deploy packages on master and if the commit is tagged. That way, we can trigger a deployment simply by creating a git tag for the release. Since it makes sense to use git tags in general to document which commits went into which version, this doesn't add any overhead. I just created a PR for JanusGraph.Net that configures Travis for this: JanusGraph/janusgraph-dotnet#11
|
@FlorianHockmann If properly configured I have no problem deploy with Travis. My issue is that @debasishdebs authentication token is hardcoded in travis.yml and will forever be public in the git history. That token needs to be setup as a secret environment variable. I was hoping @debasishdebs could redo the initial push with the credentials removed and we could just delete this branch. If he can't get to it I was figuring I could roll back my commits and then amend his initial commit to update the token to take an ENV, then reapply my PRs on top. |
I just wanted to address @debasishdebs's comment where he argued against deploying from Travis to avoid pushing multiple packages with the same version. |
|
@FlorianHockmann @chupman : Thanks for all the inputs guys. Sorry for my limited information which helped created such an argument. I didn't know about publishing from only tagged commits, but the same concept seems clear to me now. I created a new PR #29 which removed hard coded credentials (Username & PWD) in commit. It also contains a fix for #21 . What would be further steps ? Once the PR #29 is merged into |
|
Maybe once PR #29 is merged into initial branch, we create a new branch named |
|
Deleting initial branch so this is no longer required. |
Reverts #23
Reverting so credentials can be removed from .travis.yml in initial PR. Will reapply later