Dear OpenJDK Maintainers,

This bug report is a follow-up to the email discussion [1] about how OpenJDK
does not correctly delete directories on mips64el. The Bazel team discovered
this problem when the bazel-bootstrap package FTBFS on mips64el. The build
uses `deleteRecursively` from Guava which in turn should call the appropriate
system command.

From the email thread, there was a suggestion to try the mainline JDK NIO
tests. [2]

There was also a recommended patch [3] that may address this issue.

We have verified that the `deleteRecursively` functionality works correctly on
the following achitectures: amd64, arm64, ppc64el, s390x, ppc64, and riscv64.

Thanks!

So this bug is definitely not RC but would it be possible to get a fix for
it into bullseye? This seems like something that would be good to fix
before that release is stable.

Thanks in advance!

Control: tags -1 - patch
Control: tags -1+ moreinfo

Debian has porter boxes ...  please build with the proposed patch and check if
your bootstrap works.  You can run the just built jdk from the build location.

Hi Matthias,

Thanks for the suggestion! Ideally, I would love to investigate this
further. Realistically, I don't have the time (right now) to investigate a
problem in OpenJDK on an architecture I know almost nothing about. If
someone closer to the problem is able to do some troubleshooting, that
would be great! If not, I will certainly try to give it a shot myself in
the next several months.

Hi,

did anyone ever test whether the tests that failed before succeeded
with the patch applied?

I have a reason to believe that the patch may be needed but the
comment on top is incorrect:

In https://lists.zytor.com/archives/klibc/2023-January/004705.html
a problem with klibc on mips64el was pointed out leading to
https://lists.zytor.com/archives/klibc/2023-January/004706.html
which patched the arch-specific “struct stat”. If my_fstatat64_func
uses its own concept of struct stat, this may need adjusting, too.

bye,
//mirabilos