Stack traces aren't being cutoff about utest HOT 3 CLOSED

I've found out why. In BaseRunner.scala#L61, getClassName returns <jscode> for every stack trace frame.

I also tried and failed with these:

• .getClass.getName - returns java.lang.StackTraceElement for every frame.
• .getClass.getSimpleName - returns StackTraceElement for every frame.
• .getClass.getCanonicalName - not implemented in SJS.
• .getClass.getTypeName - not implemented in SJS.
• .getClass.getPackage.getName - not implemented in SJS.

from utest.

.getClassName is the right thing. It has nothing to do with getClass.getName. See StackTraceElement

However getClassName is quite unreliable. We basically try to recover this information from the VM's stack trace, which is a string with a format varying from VM to VM, and from version to version ^^. The recovery heuristics are in StackTrace. Sometimes the VM's stack trace doesn't even contain the class name, in which case there's nothing we can do.

getMethodName is pretty reliable. getFileName as well, but doesn't apply source maps in all VMs, such as PhantomJS.

The most reliable way to cut such a stack trace is to have a method with a pretty unique and recognizable name, which you can identify with getMethodName.

from utest.

This should be fixed in master by 2810c62#diff-910ec1813f88878eb7ec68c377062fb0 and the follow up f38bd2b#diff-910ec1813f88878eb7ec68c377062fb0, which uses getClassName and getMethodName to decide on stack trace cutoff, and @noinline to try and make sure Scala.js doesn't make the marker methods disappear

from utest.