RFR : 7902707 : user modules can be used only in othervm

[Igor Ignatyev]

Thank you for the reviewing and integrating the patch.

-- Igor

> On Oct 23, 2020, at 3:18 PM, Jonathan Gibbons <jonathan.gibbons at oracle.com> wrote:
> 
> Looks good to me. Approved.
> 
> Very nice!
> 
> -- Jon
> 
> On 10/16/20 11:07 AM, Igor Ignatyev wrote:
>> ping x2?
>> 
>>> On Sep 22, 2020, at 10:57 AM, Igor Ignatyev <igor.ignatyev at oracle.com> wrote:
>>> 
>>> ping?
>>> 
>>>> On Sep 17, 2020, at 2:10 PM, Igor Ignatyev <igor.ignatyev at oracle.com> wrote:
>>>> 
>>>> http://cr.openjdk.java.net/~iignatyev//CODETOOLS-7902707/webrev.00
>>>>> 230 lines changed: 175 ins; 27 del; 28 mod;
>>>> 
>>>> Hi all,
>>>> 
>>>> could you please review this patch which makes it possible to use user modules in agentvm?
>>>> from JBS:
>>>>> usage of user modules leads to CNF if a test is run in agentvm, because jtreg doesn't communicate module path to agents.
>>>>> 
>>>>> A similar problem exists in driver actions, even in othervm execution mode, as DriverAction filters out --module-path and other related flags (rerun section, though, has them listed which made it hard to comprehend)
>>>> the patch include the following changes:
>>>> - extend DO_MAIN operation message to include module-path and add-module list;
>>>> - if an agent gets a not empty modulepath, it constructs ModuleLayer is constructed using this path, and uses the loader from this layer instead of system loader;
>>>> - --module-path and --add-module are recorded in rerun section;
>>>> - for driver, --module-path, --add-modules, --add-exports, --add-opens are exempted from filtering.
>>>> 
>>>> I should notice that the driver part change is debatable, yet I'd argue the new behavior doesn't contradict the tag spec or any other official jtreg documentations and is less surprising.
>>>> 
>>>> JBS: https://bugs.openjdk.java.net/browse/CODETOOLS-7902707
>>>> webrev: http://cr.openjdk.java.net/~iignatyev//CODETOOLS-7902707/webrev.00
>>>> 
>>>> Thanks,
>>>> -- Igor