search

rax-maas / dreadnot

deploy without dread
Apache License 2.0
630 stars 61 forks source link

find oldest successful build instead of just finding the oldest #45

Closed fourk closed 11 years ago

fourk commented 11 years ago

If a successful build exists for a given revision, return the oldest successful build instead of just returning the oldest build.

pquerna commented 11 years ago

-0, if there are intermittent failures, dreadnot shouldn't support your bad decisions of moving forward.

fourk commented 11 years ago

If it is dreadnot's responsibility to restrict the user from moving forward in the case of intermittent failures, why shouldn't dreadnot fail when >0 builds fail for a given revision, as opposed to the current behavior of failing only when the oldest build resulted in failure?

On Mon, Jan 28, 2013 at 7:48 PM, Paul Querna notifications@github.comwrote:

-0, if there are intermittent failures, dreadnot shouldn't support your bad decisions of moving forward.

— Reply to this email directly or view it on GitHubhttps://github.com/racker/dreadnot/pull/45#issuecomment-12819352.

robert-chiniquy commented 11 years ago

The thing is, the only intermittent failures we're seeing are buildbot failures that are not Ele's fault. No one is going to decide to move forward with deploying an Ele build which has real issues. Currently though Buildbot will non-deterministically fail perfectly good revisions of Ele, and then that sha1 is dead to us, so people do whitespace commits and rebuild, the commit history gets polluted, and it takes too long to deploy good builds. We're working on fixing buildbot. This works around Buildbot's problems so we can deploy while we're fixing it.

dispalt commented 11 years ago

I support James solution...

-Dan

mobile: (408) 603-9137 office: (415) 779-5425

On Mon, Jan 28, 2013 at 8:07 PM, Robert Chiniquy notifications@github.comwrote:

The thing is, the only intermittent failures we're seeing are buildbot failures that are not Ele's fault. No one is going to decide to move forward with deploying an Ele build which has real issues. Currently though Buildbot will non-deterministically fail perfectly good revisions of Ele, and then that sha1 is dead to us, so people do whitespace commits and rebuild, the commit history gets polluted, and it takes too long to deploy good builds. We're working on fixing buildbot. This works around Buildbot's problems so we can deploy while we're fixing it.

— Reply to this email directly or view it on GitHubhttps://github.com/racker/dreadnot/pull/45#issuecomment-12819709.