jmx-dev [PATCH] JDK-8005472: com/sun/jmx/remote/NotificationMarshalVersions/TestSerializationMismatch.sh failed on windows

[Chris Hegarty]

On 02/08/2013 07:50 PM, Stuart Marks wrote:
> On 2/8/13 8:37 AM, Chris Hegarty wrote:
>>> Jon Gibbons suggested invoking the compiler API directly from java
>>> instead of writing a shell script. Doing this seems fairly simple, and I
>>> think it would be advantageous to keep things entirely in Java. I may
>>> attempt to rewrite the defaultSVID test using the compiler API.
>>
>> Here's a test that does just that.
>>
>> http://hg.openjdk.java.net/jdk8/tl/jdk/file/2de8c6c2d652/test/sun/misc/JarIndex/metaInfFilenames/Basic.java
>>
>
> Oh cool, interesting.
>
> This test invokes com.sun.tools.javac.Main.compile() and passes an array
> of String arguments, just like a command line. It seems pretty
> effective, at least for this case.

This com.sun API is a supported interface into the compiler. I find it 
works really well when writing tests, mainly because of it simplicity. I 
have avoided a number of shell scripts by using it.

It should be a reasonable candidate if considering removing a number of 
shell scripts, that do something trivial, then compile and run.

-Chris.

>
> But that's not what Jon was talking about. :-)
>
> I believe Jon was referring to the offical Java Compiler API (JSR 199)
> which lives in javax.tools. This API is more fiddly than just passing an
> array of strings to javac's main, but it's also considerably more
> flexible. For example, it's possible to use it to call the compiler
> directly on a String! (I haven't coerced it to compile a String directly
> to class bytes, but it should be possible to do so in principle.)
>
> Anyway, this is good, we should investigate these approaches for a
> variety of use cases and develop a repertoire of techniques. Eventually
> we may be able to avoid writing shell script tests entirely.
>
> s'marks
>
>