RFR: 8195129: System.load() fails to load from unicode paths [v3]

[Maxim Kartashev]

On Mon, 7 Jun 2021 18:46:11 GMT, Naoto Sato <naoto at openjdk.org> wrote:

>> @mkartashev  thank you for the detailed explanation.
>> 
>> It is not clear to me that the JDK's conformance to being a Unicode application has significantly changed since the evaluation of JDK-8017274 - @naotoj  can you comment on that and related discussion from the CCC for JDK-4958170 ? In particular I'm not sure that using the platform encoding is wrong, nor how we can have a path that cannot be represented by the platform encoding?
>> 
>> Not being an expert in this area I cannot evaluate the affects of these shared code changes on other platforms, and so am reluctant to introduce any change that affects any non-Windows platforms. Also the JVM and JNI work with modified-UTF8 so I do not think we should diverge from that.
>> I would hate to see windows specific code introduced into the JDK or JVM's shared code for these APIs, but that may be the only choice to avoid potential disruption to other platforms. Though perhaps we could push the initial conversion down into the JVM?
>
> @dholmes-ora Sorry, I don't think anything has changed as to the encoding as of JDK-8017274. For some reason, I had the impression that JVM_LoadLibrary() accepts UTF-8 (either modified or standard), but that was not correct. It is using the platform encoded string for the pathname.
> 
> @mkartashev As you mentioned in another comment, the only way to fix this issue is to pass UTF-8 down to JVM_LoadLibray, but I don't think it is feasible. One reason is the effort is too great, and the other is that all VM implementations would need to be modified.

@naotoj Then I guess this bug will have to wait until Windows evolves to the point when its platform encoding is UTF-8. In the mean time, I'm closing this PR. 

Thank you all so much for your time!

-------------

PR: https://git.openjdk.java.net/jdk/pull/4169