release command

[eed3si9n]

Problem

sbt 1.7.0 made ++ command more strict, making it not possible
to use +publishSigned as is.
This is because compiler plugin has much finer-grained
crossScalaVersions compared to the core libraries.

Solution

Make all subprojects as fine-grained as needed by the compiler plugins.
Adjust publishing by defining a custom command called release.

Note: CI_RELEASE / CI_SNAPSHOT_RELEASE need to be adjusted accordingly.

[tgodzik]

Thanks @eed3si9n ! I updated the release.yml to use the new command when releasing. I think this should work, though I wonder if there is anything that we can do to handle such situations by default in sbt?

Or maybe the use case here is super niche and quite legacy, so ultimately it's not a project structure we would want to recommend?

[tgodzik]

@eed3si9n I am trying to run the release, but it seems that each time it hangs https://github.com/scalameta/scalameta/runs/7433154721?check_suite_focus=true

Looks like memory issues, which we didn't encounter before. Each time the version changes it seems to take on more and more memory, probably due to some caching? Anyway to force sbt to free some memory?

[eed3si9n]

I suggest you give it more memory. For example, this is what we use for sbt - https://github.com/sbt/sbt/blob/1.7.x/.github/workflows/ci.yml:

    env:
      JAVA_OPTS: -Xms800M -Xmx2G -Xss6M -XX:ReservedCodeCacheSize=128M -server -Dsbt.io.virtual=false -Dfile.encoding=UTF-8
      JVM_OPTS: -Xms800M -Xmx2G -Xss6M -XX:ReservedCodeCacheSize=128M -server -Dsbt.io.virtual=false -Dfile.encoding=UTF-8

[tgodzik]

I suggest you give it more memory. For example, this is what we use for sbt - https://github.com/sbt/sbt/blob/1.7.x/.github/workflows/ci.yml:

    env:
      JAVA_OPTS: -Xms800M -Xmx2G -Xss6M -XX:ReservedCodeCacheSize=128M -server -Dsbt.io.virtual=false -Dfile.encoding=UTF-8
      JVM_OPTS: -Xms800M -Xmx2G -Xss6M -XX:ReservedCodeCacheSize=128M -server -Dsbt.io.virtual=false -Dfile.encoding=UTF-8

We have even higher memory set and it's still running out of memoery. I thought that it might have been the amount of classes being loaded up, so I increased -XX:ReservedCodeCacheSize but that didn't help.

Getting a lot of:

[warn] In the last 16 seconds, 8.898 (55.9%) were spent in GC. [Heap: 2.94GB free of 5.21GB, max 5.33GB] Consider increasing the JVM heap using `-Xmx` or try a different collector, e.g. `-XX:+UseG1GC`, for better performance.
[warn] one warning found

which is weird since it seems to have a lot of free memory. Why would it GC so heavily?