The free_callback function has a JNI weak global reference leak problem about jna HOT 6 CLOSED

Thank you. I think I see the reasoning in the structure. However, could you pleas check the example again? When I run it, I immediately get a segfauls after the first iteration. This happens for JDKs 11, 17 and 21.

from jna.

With this additional change:

--- a/src/com/sun/jna/CallbackReference.java
+++ b/src/com/sun/jna/CallbackReference.java
@@ -508,7 +508,7 @@
         Map<Callback, CallbackReference> map = direct ? directCallbackMap : callbackMap;
         synchronized(pointerCallbackMap) {
             CallbackReference cbref = map.get(cb);
-            if (cbref == null) {
+            if (cbref == null || cbref.cbstruct == null) {
                 cbref = new CallbackReference(cb, callingConvention, direct);
                 map.put(cb, cbref);
                 pointerCallbackMap.put(cbref.getTrampoline(),

I can run the sample in a stable manner and reproduce the problem. In my test I see 700MB RSS vs. 385MB RSS with 10.000.000 iterations.

One final question @xiezhaokun: Could you check your author information: xiezhaokun <[email protected]>. Is xiezhaokun really your full name (given and familyname) and is the email a valid email address?

from jna.

With this additional change:

--- a/src/com/sun/jna/CallbackReference.java
+++ b/src/com/sun/jna/CallbackReference.java
@@ -508,7 +508,7 @@
         Map<Callback, CallbackReference> map = direct ? directCallbackMap : callbackMap;
         synchronized(pointerCallbackMap) {
             CallbackReference cbref = map.get(cb);
-            if (cbref == null) {
+            if (cbref == null || cbref.cbstruct == null) {
                 cbref = new CallbackReference(cb, callingConvention, direct);
                 map.put(cb, cbref);
                 pointerCallbackMap.put(cbref.getTrampoline(),

I can run the sample in a stable manner and reproduce the problem. In my test I see 700MB RSS vs. 385MB RSS with 10.000.000 iterations.

One final question @xiezhaokun: Could you check your author information: xiezhaokun <[email protected]>. Is xiezhaokun really your full name (given and familyname) and is the email a valid email address?

Yes, my user information is all valid.

from jna.

Thank you. I think I see the reasoning in the structure. However, could you pleas check the example again? When I run it, I immediately get a segfauls after the first iteration. This happens for JDKs 11, 17 and 21.

The test case I used jna 5.14.0 version to test

from jna.

Thank you. I think I see the reasoning in the structure. However, could you pleas check the example again? When I run it, I immediately get a segfauls after the first iteration. This happens for JDKs 11, 17 and 21.

I updated the example.

from jna.

Fixed via 4bf9f92

from jna.