Adam Murdoch wrote > On 14/11/2012, at 2:30 PM, Peter Niederwieser wrote: >> the combination of `useAnt=false` (i.e. Zinc) and `fork=false` doesn't >> work. I implemented this because there was a concrete need to have (or >> keep) >> each of the other three combinations, and it felt a bit strange to leave >> out >> the fourth one. I thought that by reusing the >> InProcessCompilerDaemonFactory >> we have for Java and Groovy, it would be trivial to implement this. >> Unfortunately, I forgot to add an integration test for this case. (We are >> testing quite a few combinations by now.) As it turns out, the current >> InProcessCompilerDaemonFactory has performance and correctness problems >> when >> used with Scala. > > What are the issues?
I haven't investigated all test failures, but one cause I've identified is class loader inconsistencies. For example, `catch(CompilationFailedException)` in NormalizingScalaCompiler doesn't catch anything because the `CompilationFailedException` class thrown has a different defining class loader. Another potential problem is that the cache loaders aren't cached, so I assume we'll incur the same performance penalties as back when we didn't use a compiler daemon. Cheers, Peter -- View this message in context: http://gradle.1045684.n5.nabble.com/Known-problems-with-new-Scala-features-in-1-3-rc-1-tp5710401p5710409.html Sent from the gradle-dev mailing list archive at Nabble.com. --------------------------------------------------------------------- To unsubscribe from this list, please visit: http://xircles.codehaus.org/manage_email
