<i18n dev> RFR: 8158272 & 8158468 (tools/jlink/plugins/IncludeLocalesPluginTest.java bug fixes)

[Alan Bateman]

On 10/06/2016 08:08, Masayoshi Okutsu wrote:
> (re-sending to include jigsaw-dev)
>
> Hi,
>
> Please review fixes for 8158272 and 8158468. The test had several 
> problems.
>
> - A child process doesn't inherit IO. Any outputs from the child 
> process are not logged.
> - The exit code of a child process is ignored. The exit code needs to 
> be checked by the test.
> - A child process should use the exit code to report errors rather 
> than throwing a RuntimeException.
> - The golden data doesn't match the CLDR V29.
> - It may not be a good idea to hard-code the full set of the available 
> locales.
>
> All have been fixed. Additional changes are:
>
> - Removed -verbose:gc from @run
> - Changed String to List<String> for the available locales data. It 
> was hard to compare long strings.
> - Changed output of the test so that it's easier to understand test 
> activities.
> - Replaced Locale.ROOT.toString() with "(root)"
>
> As for the timeout issue, what happened is that a child process for 
> checking available locales threw a RuntimeException due to the CLDR 
> V29 changes. But the parent process (the main test process) was 
> waiting for the child process to terminate. I'm not sure if it's a 
> Windows specific process management issue or jtreg specific issue or 
> something else. But the timeout symptom is no longer reproducible with 
> the change to call ProcessBuilder.inheritIO().
>
> Issues:
> https://bugs.openjdk.java.net/browse/JDK-8158272
> https://bugs.openjdk.java.net/browse/JDK-8158468
>
> Webrev:
> http://cr.openjdk.java.net/~okutsu/9/8158272.8158468/webrev.00
>
> While I was working on these test problems, I noticed there are a few 
> problems in the include locales plugin per se. For example, 
> --include-locales=*-IN selects en-001 which is missing in the 
> available locales list. (The missing "en-001" has been added to the 
> test data, but it's commented out.) I will be filing a separate JBS 
> issue.
This looks quite good.

What would you think also changing testAvailableLocales to use 
ProcessTools. As it stands then it looks like the output/error streams 
from the child won't be printed to the test logs?

-Alan