RFR: 8232864: Classes generated by GenerateJLIClassesPlugin.java are not reproducable

[Jie Fu]

Thanks Claes for your review and comments.

On 2019/10/23 下午6:49, Claes Redestad wrote:
> Hi,
>
> ideally the build step to generate default_jli_trace.txt should be
> done independently of where the CDS feature is compiled in, but the
> fix looks good regardless.
>
> /Claes
>
> On 2019-10-23 09:16, Jie Fu wrote:
>> Hi all,
>>
>> JBS:    https://bugs.openjdk.java.net/browse/JDK-8232864
>> Webrev: http://cr.openjdk.java.net/~jiefu/8232864/webrev.00/
>>
>> To find out how to reproduce this bug, please refer to the JBS.
>>
>> To keep things sorted in a deterministic order, current 
>> implementation seems to be dependent on a default trace file [1] 
>> (default_jli_trace.txt, which is generated during building the jdk 
>> image [2]).
>> However, not all images will generate such a trace file.
>> For example, when cds is disabled, default_jli_trace.txt won't be 
>> generated at all.
>> If the trace file doesn't exist, classes generated by 
>> GenerateJLIClassesPlugin.java will not be sorted in a deterministic 
>> order.
>>
>> Testing:
>>   - make test TEST="tier1 tier2 tier3" CONF=server-release
>>
>> Could you please review it and give me some advice?
>>
>> Thanks a lot.
>> Best regards,
>> Jie
>>
>> [1] 
>> http://hg.openjdk.java.net/jdk/jdk/file/70e6b0d8db13/src/jdk.jlink/share/classes/jdk/tools/jlink/internal/plugins/GenerateJLIClassesPlugin.java#l58 
>>
>> [2] 
>> http://hg.openjdk.java.net/jdk/jdk/file/70e6b0d8db13/make/GenerateLinkOptData.gmk#l54 
>>
>>
>>