add sbt-reproducible-builds and fix lanucher builds

[Roiocam]

Continue on the #205, resolves: #192

Jar results seem the same:

❯ diff unzip.txt unzip_main.txt
1,2c1,2
< ❯ unzip pekko-persistence-cassandra-launcher_2.13.jar
< Archive:  pekko-persistence-cassandra-launcher_2.13.jar
---
> ❯ unzip pekko-persistence-cassandra-launcher_2.13_main.jar
> Archive:  pekko-persistence-cassandra-launcher_2.13_main.jar
4,6d3
<   inflating: META-INF/LICENSE
<   inflating: META-INF/NOTICE
<   inflating: cassandra-bundle.jar
12a10,12
>   inflating: META-INF/LICENSE
>   inflating: META-INF/NOTICE
>   inflating: cassandra-bundle.jar

and the zipdiff report the same result:

 java -jar ~/Downloads/zipdiff-0.4/build/zipdiff.jar -file1 ./pekko-persistence-cassandra-launcher_2.13.jar -file2 main/pekko-persistence-cassandra-launcher_2.13_main.jar -outputfile diffs.html

❯ diff META-INF/MANIFEST.MF main/META-INF/MANIFEST.MF
3c3,5
< Implementation-URL: https://pekko.apache.org/
---
> Implementation-Version: 1.1.0-M0+63-fa9370e2-SNAPSHOT
> Specification-Vendor: Apache Software Foundation
> Specification-Title: pekko-persistence-cassandra-launcher
4a7,8
> Specification-Version: 1.1.0-M0+63-fa9370e2-SNAPSHOT
> Implementation-URL: https://pekko.apache.org/
6d9
< Implementation-Version: 1.1.0-M0+67-c7590392-SNAPSHOT
9,11d11
< Specification-Title: pekko-persistence-cassandra-launcher
< Specification-Vendor: Apache Software Foundation
< Specification-Version: 1.1.0-M0+67-c7590392-SNAPSHOT

[Roiocam]

This is PR change, can you take a look? thanks @raboof

[raboof]

This does make sense as as those build results are 'unmanaged' from the perspective of the cassandra-launcher build.

I might worry sbt wouldn't correctly pick up that changes to the bundle should trigger a rebuild of the launcher, but perhaps that's sufficiently guaranteed since cassandraBundle / Compile / packageBin is still part of the graph.

In any case I don't think we expect many changes here, and CI shows it works well also when building 'from scratch'. I've also tried various experiments in sbt and didn't see any new/seriously problematic behavior.

For reference: I did notice that sbt clean does not clear cassandra-bundle/target, and it does not get rebuilt when it exists and you're building sbt cassandraLauncher/package, even if you made changes to the definition of the bundler. That was all already the case on the main branch, though, so not a regression.

[He-Pin]

lgtm

[raboof]

thanks a lot for digging into this!

[raboof]

This does make sense as as those build results are 'unmanaged' from the perspective of the cassandra-launcher build.

I might worry sbt wouldn't correctly pick up that changes to the bundle should trigger a rebuild of the launcher, but perhaps that's sufficiently guaranteed since cassandraBundle / Compile / packageBin is still part of the graph.

In any case I don't think we expect many changes here, and CI shows it works well also when building 'from scratch'. I've also tried various experiments in sbt and didn't see any new/seriously problematic behavior.

For reference: I did notice that sbt clean does not clear cassandra-bundle/target, and it does not get rebuilt when it exists and you're building sbt cassandraLauncher/package, even if you made changes to the definition of the bundler. That was all already the case on the main branch, though, so not a regression.