I think we have a problem with the way e2e tests are currently structured.
They run all components in a single JVM - a setup that is never used in
practice. As a result, this creates a lot of potential for breaking things
due to various library version issues. I've seen a few cases where
reasonable changes to Sentry break because of this library version
incompatibility in the combined VM image.

I think 2e tests should be restructured to resemble the way things are
running in real life - as separate Java processes, so that there are no
weird test issues caused by combining them in a single JVM.

What do people think about this?

- Alex.

Reply via email to