RFR: 8352728: InternalError loading java.security due to Windows parent folder permissions [v11]

[Francisco Ferrari Bihurriet]

On Wed, 26 Nov 2025 21:17:42 GMT, Sean Mullan <mullan at openjdk.org> wrote:

>> Hi folks, as we know [JDK 26 enters Rampdown Phase One next week](https://mail.openjdk.org/pipermail/jdk-dev/2025-November/010639.html) on Thursday, 4 December.
>> 
>> Please let me know if we'll want this fix to move forward before that happens, and please note that soon I will need to focus on the OJVG work.
>> 
>> If you prefer, I'm fine with using `path.toRealPath()` instead of `path.toFile().getCanonicalFile().toPath()`. That would keep relative includes broken in _Windows_ UWP Java applications and from _Linux_ anonymous pipes and files (where they don't make a lot of sense anyway). But that is better than the current status quo (_Windows_ UWP Java applications break if they load the `java.security.Security` class).
>> 
>> Just let me know if you want to proceed that way and I will, also removing the part of the new test that will break by that change.
>
>> Hi folks, as we know [JDK 26 enters Rampdown Phase One next week](https://mail.openjdk.org/pipermail/jdk-dev/2025-November/010639.html) on Thursday, 4 December.
>> 
>> Please let me know if we'll want this fix to move forward before that happens, and please note that soon I will need to focus on the OJVG work.
>> 
>> If you prefer, I'm fine with using `path.toRealPath()` instead of `path.toFile().getCanonicalFile().toPath()`. That would keep relative includes broken in _Windows_ UWP Java applications and from _Linux_ anonymous pipes and files (where they don't make a lot of sense anyway). But that is better than the current status quo (_Windows_ UWP Java applications break if they load the `java.security.Security` class).
>> 
>> Just let me know if you want to proceed that way and I will, also removing the part of the new test that will break by that change.
> 
> Yes, I agree that seems like a way forward, if not perfect. @AlanBateman ok with you?

Hi @seanjmullan, after a sync-up with `master`, I needed to go back to `path.toRealPath()` anyway (details in my [previous comment](#issuecomment-3582665183)), so I think the only pending discussions are these two:

* [Inline `Files.isRegularFile()` call and remove I/O saving shortcut to skip `Files.isDirectory()`?](#discussion_r2519385066)
* [Re-wrap `IOException` thrown by `Files.isSameFile()` as `InternalError` and stop ignoring it?](#discussion_r2519491637)

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

PR Comment: https://git.openjdk.org/jdk/pull/24465#issuecomment-3583601011