I need to provide picard. It's a java application that depends on at least one other library (htdjdk). Building from source seems opaque while just plopping the jar file into prefix seems easy. I wouldn't suggest using spack to copy a platform-specific binary into place, but since jar files are platform agnostic, perhaps this isn't so offensive. Or, perhaps it is.

What's the Spack stand on copying jar files from upstream providers?

Additional details....

I haven't totally sussed out the build from source bit, but it seems like it's going to need the htsjdk jar and is likely to reach out and grab it. There are instructions for using one's own htsjdk.jar, but they include this bit describing what to do with that jar

to install that version to your local maven repository

All of this seems like a lot of overhead....

My users would be content if I grabbed the distributed jar file, dropped it into prefix and set the $PICARD environment variable to point there.

Picard's instructions recommend running it as follows:

java -jar /path/to/picard.jar -h

and their warm fuzzy recommendation is to abstract that path into an environment variable:

java -jar $PICARD -h 

with the general usage pattern being:

java jvm-args -jar $PICARD PicardToolName OPTION1=value1 OPTION2=value2...

Apparently no one has revolted and written a picard bash script to wrap it up.....