Request for approval for 7074189

[Jonathan Gibbons]

Bug: 7074189

JDK 8 Changeset: 
http://hg.openjdk.java.net/jdk8/tl/langtools/rev/c0d5f93af048

The bug is not currently visible on bugs.sun.com.  Here is the relevant 
info:

*Synopsis*: some javac tests fail with latest jtreg 4.1 b03

===*Description*  ============================================================
3 tests fail using jtreg 4.1 b03

===*Evaluation*  =============================================================
There are three test failures:

tools/javac/processing/errors/TestOptionSyntaxErrors.java
tools/javac/processing/errors/TestReturnCode.java
tools/javac/warnings/Serial.java

*** (#1 of 4): 2011-08-03 17:03:10 GMT+00:00jonathan.gibbons at oracle.com

1. test/tools/javac/warnings/Serial.java

The action
  * @compile/fail -Werror -Xlint:all,-path T4994049/ Serial.java
is nonsensical, and succeeding because the compilation fails for the wrong reason.

The line should be deleted.

*** (#2 of 4): 2011-08-03 17:03:10 GMT+00:00jonathan.gibbons at oracle.com

2.  The other two tests are expecting "less common" exit codes from javac. The best way to fix them is to write a library class, CompileFail, which emulates @compile/fail but which also allows the specific exit code to be tested.

*** (#3 of 4): 2011-08-03 22:31:37 GMT+00:00jonathan.gibbons at oracle.com

Hmmm, we've uncovered a curiousity in TestReturnCode.java.   Errors thrown from annotation processors are treated different (and less severely) than exceptions.

Errors cause EXIT_ERROR, exit code 1, equivalent to a compile-time error

Exceptions cause EXIT_SYSERR, exit code 3, system error or resource exhaustion

*** (#4 of 4): 2011-08-03 22:43:58 GMT+00:00jonathan.gibbons at oracle.com