GH-759: Get length of byte[] in TryCopyLastError

[hnwyllmm]

What's Changed

Please fill in a description of the changes here.

We should get the length of byte[] by GetArrayLength but not strlen which may cause invalid memory access.

Closes #759 .

[lidavidm]

Is it possible to add a test?

[hnwyllmm]

The error message in CI doesn't seem to be related to my PR:

/opt/rh/devtoolset-10/root/usr/libexec/gcc/x86_64-redhat-linux/10/ld: cannot find -lutf8proc
  collect2: error: ld returned 1 exit status

linkage: https://github.com/apache/arrow-java/actions/runs/15140171347/job/42594091462?pr=760

Do you have some suggestions?

[hnwyllmm]

Is it possible to add a test?

I'll try to add a test case in integration_tests.py.

[kou]

The error message in CI doesn't seem to be related to my PR:

/opt/rh/devtoolset-10/root/usr/libexec/gcc/x86_64-redhat-linux/10/ld: cannot find -lutf8proc
  collect2: error: ld returned 1 exit status

linkage: https://github.com/apache/arrow-java/actions/runs/15140171347/job/42594091462?pr=760

Do you have some suggestions?

Could you open a separated issue for it?
Let's fix it as a separated task.

[lidavidm]

Is it possible to add a test?

I'll try to add a test case in integration_tests.py.

Do you need an integration test? IMO, you should be able to export a stream, then immediately import it, all within Java.

[hnwyllmm]

The error message in CI doesn't seem to be related to my PR:

/opt/rh/devtoolset-10/root/usr/libexec/gcc/x86_64-redhat-linux/10/ld: cannot find -lutf8proc
  collect2: error: ld returned 1 exit status

linkage: https://github.com/apache/arrow-java/actions/runs/15140171347/job/42594091462?pr=760
Do you have some suggestions?

Could you open a separated issue for it? Let's fix it as a separated task.

issue: #767

[kou]

Thanks. (We've fixed this problem in apache/arrow. Sorry for not sharing it here...)

[hnwyllmm]

Is it possible to add a test?

I'll try to add a test case in integration_tests.py.

Do you need an integration test? IMO, you should be able to export a stream, then immediately import it, all within Java.

I add a unittest but I think it can't make sure TryCopyLastError works correctly, so I suggest it should be removed.
We can read memory after the error string address in most cases. We usually check such problems by sanity tools.

[lidavidm]

Can't the test validate that the exception has the right error message after making it through the FFI boundary?