Expose the generated rpm filepath

The generated rpm filename should be exposed via a setting.

I generate an rpm via sbt rpm:packageBin in a CI which I then upload to a nexus using curl:

# This generates the rpm but does not log anything to the console
PACKAGE_OUTPUT=$(sbt "print rpm:packageBin") 
echo $PACKAGE_OUTPUT
# Extract the rpm filename full path
RPM_FILE=$(echo $PACKAGE_OUTPUT | tail -n1)
curl --user "$NEXUS_USERNAME:$NEXUS_PASSWORD" --upload-file $RPM_FILE https://nexus.example.com/repository/yum-releases/7/os/$ARCH/$(basename $RPM_FILE)

But its quite clunky as the sbt "print rpm:packageBin" is what actually generates the rpm. It is also the only way to know the generated rpm filename (apart from computing it from the other settings). The code that would be easy to read should be something like:

sbt "rpm:packageBin"
RPM_FILE=$(sbt "rpm:filename" | tail -n1)
curl --user "$NEXUS_USERNAME:$NEXUS_PASSWORD" --upload-file $RPM_FILE https://nexus.example.com/repository/yum-releases/7/os/$ARCH/$(basename $RPM_FILE)

Information

• What sbt-native-packager are you using: 1.4.1
• What sbt version: 1.3.2
• What is your build system: Arch Linux + rpm-org package
• What package are you building: rpm
• What version has your build tool: 4.15.0
• What is your target system: CentOS 7

[Lasering]

@muuki88 yes!

[nigredo-tori]

@muuki88,

returns only the version and nothing else

This also means that we lose most of the log in case anything goes wrong. Also, I don’t think this would work if some task logs something at the error level.

I would argue that the RPM path is well known as well. Not the exact file, but the output directory which only contains one file.

1. The almost known path is not always convenient. E.g. I have a hand-written Dockerfile that I run after a build (I haven’t switched to DockerPlugin yet), so without a setting like this I would have to either postprocess the build output to move the file to a known path (easy, but annoying), or pass this uncertainty to the resulting docker image.
2. There is no guarantee that there is only one file, unless the process is happening in CI. E.g. if I package, change version and package again, I will probably get two files. Again, this can be mitigated by cleaning the output folder before the build - but that is also annoying.

I consider this a hack as directory != file

That, indeed, would be a hack. Which is why we should use artifactPath instead:

Rpm / packageBin / artifactPath := ???
Rpm / packageBin := {
   val rpmFile := (Rpm / packageBin / artifactPath).value
}

This also doesn’t require any new settings, but it’s as idiomatic as it gets - you’ll find that Compile / packageBin uses Compile / packageBin / artifactPath for the same purpose, albeit with a bit of indirection.