On Wed, 14 Jan 2026 22:46:58 GMT, Chris Plummer <[email protected]> wrote:
>> Thanks, added.
>
> On windows-x64 I'm seeing about 25
> NoFramePointer!Java_LingeredAppWithNativeMethod_callJNI frames. On other
> platforms I see just one Java_LingeredAppWithNativeMethod_callJNI frame, and
> a couple of fib() frames above it. Maybe this difference is just native
> compiler inlining and tail call elimination, but wanted to point it out just
> to make sure it is correct.
On my Windows 11 25H2, mixed jstack works correctly like following:
<snip>
0x00007ff8105c10c4 NoFramePointer!fib + 0x28
0x00007ff8105c10c4 NoFramePointer!fib + 0x28
0x00007ff8105c101d NoFramePointer!Java_LingeredAppWithNativeMethod_callJNI
+ 0x1d
0x000001c5903afbf8 * LingeredAppWithNativeMethod.callJNI(java.lang.Object,
int) bci:0 (Interpreted frame)
* LingeredAppWithNativeMethod.callNative() bci:47
line:54 (Interpreted frame)
* LingeredAppWithNativeMethod.callback(long) bci:8
line:63 (Interpreted frame)
<snip>
Do you mean multiple jstacks are shown in .jtr file with message such as
"DEBUG: Iteration: 1 - Test didn't trigger interesting condition." ? If yes, I
think it is expected.
TestJhsdbJstackMixed.java seems to check at `isFibAndAlignedAddress` whether
instruction address (PC) is aligned or not. I saw the condition is met in early
on Linux x64, but it seems to take some calls on Windows x64. I'm not sure why
we need to check this (I checked
[JDK-8208091](https://bugs.openjdk.org/browse/JDK-8208091) which introduced it,
but I haven't yet understood the reason).
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/29019#discussion_r2692562088
