Il giorno lunedì 21 gennaio 2013 14:24:25 UTC+1, szabolcs.berecz ha scritto:
Maybe scalac is printing to stderr in this case?
Nope, it's using Console.println:
protected def explain[T](op: String, p: (Type, T) => Boolean, tp1: Type, arg2: T): Boolean = {
Console.println(indent + tp1 + " " + op + " " + arg2 + "?" /* + "("+tp1.getClass+","+arg2.getClass+")"*/)
and I remember clearly adding debug println statements without failing tests.
I don't know what's going on here exactly, but I started working on SI-6289 a couple weeks ago which is mostly about ignored stderr output. Currently, anything printed to stderr is ignored.
That would be very bad, but luckily that's only for javac. I just saw this failure from partest on master, and the output is on stderr.
testing: [...]/files/neg/migration28.scala [FAILED]
1d0
< warning: -Xmigration is deprecated: This setting is no longer useful and will be removed. Please remove it from your build.
7c6
< two warnings found
---
> one warning found
Also, while working on this bug I noticed that a lot of stdout output is also ignored by partest.
Yes, that's more like it.
It's mostly warnings and deprecation messages but there might be others, I haven't checked everything. Running "partest --update-checks" adds 1-2 thousand lines to the check files. What's causing it: the log file is passed around as a File in partest and it's almost never opened in append mode. So, this can also be the cause of your issues.
That should also be filed under JIRA, in the same or another bug. Can you do that, since you have more clue?
Also, I reproduced the problem with the test and raised the priority of the bug to blocker. Doesn't that mean that tests for compatibility with Java are being ignored? That would be bad. Strictly speaking, no other change whatsoever should be merged until such a bug is fixed.
Or is the bug specific to executing individual tests? In that case (which would be weird), I apologize for the fuss.