RFR : 8016577 : (s) langtools/test/Makefile should only set some vars if unset

[Jonathan Gibbons]

Mike,

Thanks for looking at this.

I guess I am not a big fan of JT_ as an abbreviation of JTREG_.

Lines 149-151 seem off-color to me.  Based on the name, I would never 
expect $TESTJAVA to contain dist/lib/classes.jar.    The previous code 
looked for distlib/classes.jaar in $PRODUCT_HOME which was a more 
accurate reflection of what was going on.

-- Jon


On 06/13/2013 01:11 PM, Mike Duigou wrote:
> Hello all;
>
> I have been working on improving the execution of tests via the root repo "make test" target (JDK-8009683). This work has so far mostly concerned the jdk repo but now I would like to increase the commonality of the test/Makefile for the various repos.
>
> In the langtools/test/Makefile these vars are
>
> JTREG_HOME
> JT_JAVA
> TESTJAVA
>
> currently unconditionally set. They should only be set if they don't already have definitions. This is part of an effort to retire some of the JPRT specific vars and synchronize the jdk and langtools test/Makefile
>
> Additionally
>
> - JTREG_HOME is renamed to JT_HOME for consistency
>
> - TESTBOOTCLASSPATH setting is made more uniform, ie. not just when PRODUCT_HOME is defined.
>
> - Warnings are now printed for variables using the default "unset" value. In most cases these variable *should* already have a value so no default should be used. For now the defaults are retained but with a warning.
>
> - An error is generated if PRODUCT_HOME seems to point at an invalid location.
>
> A review webrev is posted here:
>
> http://cr.openjdk.java.net/~mduigou/JDK-8016577/0/
>
> Once this issue is completed I will follow up with JDK-8016573 which will replace JPRT_JAVA_HOME with JT_JAVA in the Main.gmk make file.
>
> Cheers,
>
> Mike