Use hashing instead of other algorithms to guarantee scala classes re…

[fido-node]

…build.
Fix for #307

[thesamet]

Thanks for sending the PR. Since getting the hash is more expensive, I am wondering why the "last modified" approach fails here? Do the files in the two branches have the exact same last modified time?

[fido-node]

Hi. It is a little bit hard for me to debug why this happens.
I've attached reproduction case and repo with the smallest possible example to this issue:
#307

[fido-node]

This behavior related to this enhancement in sbt. Since our build pipelines do not really need this sbt feature I can close this PR for now. I'll also write a comment in the issue to describe proper workaround.

[thesamet]

Thanks for reporting, investigating and documenting the workaround!

[bjaglin]

I don't have time to test nor to write a scripted test, but I believe this should address the issue with no overhead (actually it would save some FS calls).

sbt-protoc/src/main/scala/sbtprotoc/ProtocPlugin.scala

Line 491 in 000df10

val set = IO.unzip(dep, extractTarget, "*.proto")

-val set = IO.unzip(dep, extractTarget, "*.proto")
+val set = IO.unzip(dep, extractTarget, "*.proto", preserveLastModified = false)