Closed pwendell closed 10 years ago
@jsuereth any thoughts on this one?
Hmm.... I was hoping maybe there would be another way, but perhaps this is the best.
Thanks for the patch! Do you need a release with this?
Hey @jsuereth totally open to finding a better way - I just didn't see an obvious one.
From the Spark side there are a couple other features that we need so I wouldn't bother spinning a release yet. One is being able to load profiles and the other is being able to pass user properties. The plumbing for this in maven is super simple (you just need to call some functions when loading the pom). Mostly I'm just still learning how to add configuration options correctly in sbt.
cool.
Word of warning: Please don't publish artifacts using profiles. I'm pretty thoroughly convinced it's a great evil that has been inflicted on us. If you do use profiles, it'd be nice if you could publish using an 'effective-pom' so that non-maven tools aren't completely hosed trying to consume artifacts.
We rely on profiles mostly packaging different types of binary releases. They aren't intended for downstream consumption - i.e. any downstream consumption can just ignore the profiles and be fine.
On Wed, Mar 26, 2014 at 3:15 PM, Josh Suereth notifications@github.comwrote:
cool.
Word of warning: Please don't publish artifacts using profiles. I'm pretty thoroughly convinced it's a great evil that has been inflicted on us. If you do use profiles, it'd be nice if you could publish using an 'effective-pom' so that non-maven tools aren't completely hosed trying to consume artifacts.
Reply to this email directly or view it on GitHubhttps://github.com/sbt/sbt-pom-reader/pull/4#issuecomment-38747308 .
Unfortunately thought we do publish the profiles because our publishing goes through Maven not sbt due to using the Apache release plug-in. It might be nice to see if we could use sbt for publishing in the future for this reason.
On Wed, Mar 26, 2014 at 3:25 PM, Patrick Wendell pwendell@gmail.com wrote:
We rely on profiles mostly packaging different types of binary releases. They aren't intended for downstream consumption - i.e. any downstream consumption can just ignore the profiles and be fine.
On Wed, Mar 26, 2014 at 3:15 PM, Josh Suereth notifications@github.comwrote:
cool.
Word of warning: Please don't publish artifacts using profiles. I'm pretty thoroughly convinced it's a great evil that has been inflicted on us. If you do use profiles, it'd be nice if you could publish using an 'effective-pom' so that non-maven tools aren't completely hosed trying to consume artifacts.
Reply to this email directly or view it on GitHubhttps://github.com/sbt/sbt-pom-reader/pull/4#issuecomment-38747308 .
Anything I can do, in general, to help with sbt issues you have, or help fit it into the apache structure, let me know. We're more than happy to make sure your experience is smooth or fix whatever gaps might exist.
I'm guessing the Apache release plugin allows maven projects to release into Apache infrastructure?
The approach of guessing the directory name doesn't work so well for us in Spark because we have a fairly complicated nested set of projects.
This proposal allows the user to add a Maven property
sbt.project.name
which overrides the directory name.