rfm

Any update on a way to reproduce this issue @eukara ?

I haven't set up the appropriate NFS to try it out yet, but looking at the description of the test case and the code I can see two clear issues...

Failing to set correct ownership etc is often a major security issue, so yes there is at least one huge reason why the operation should fail (and there may be...

I have reproduced this ... the underlying issue is that the directory can't be fully copied because the nfs mount doesn't permit the ownership to be set to the current...

But, at one point in the code, the result of calling -fileManager:shouldProceedAfterError: was ignored. That *is* a bug. This has no effect in the testcase (which fails to set a...

PS. thanks for the testcase info, particularly the nfs setup and the ideas about likely reasons for copying the attributes to be impossible. Without that information the issue would likely...

The way it's supposed to work is that clang and libobjc2 go together, and this is indicated by the use of 'ng' rather than 'gnu' in the library combo, so...

I detest the bloated nature of junit, but I do recognise that it's widely used/supported, and that feeding test results into external tools could be useful, so I don't see...

A struct return issue perhaps? Maybe in the ffi library or maybe in the way it's used? > > On 5 Nov 2025 at 13:52, Riccardo ***@***.***> wrote: > >...

It looks like the crash is because v is nil. I think you could have a race condition on program exit ... if +[NSValue atExit:] is called before this code...