Hi Daniil, It'd be nice if Chris has a chance to look at this fix. He has most recent experience in this area.
Thanks, Serguei On 1/17/19 6:08 PM, Daniil Titov wrote:
Please review the change that fixes JDB stepping issue for a specific case when the single step request was initiated earlier in the stack, previous calls were for methods in the filtered classes (single stepping was disabled), handleMethodEnterEvent() re-enabled stepping and the first bytecode upon entering the current method requires resolving constant pool entry. In this case the execution resumes in java.lang.Classloader.loadClass() and since it is also a filtered class the single stepping is getting disabled again (stepControl.c :593). When loadClass() exits a notifyFramePop() is called on the loadClass() frame but due to condition fromDepth >= afterPopDepth at stepControl.c :346 (that doesn't hold in this case, in this case fromDepth is 1 and afterPopDepth is 4) the notifyFramePop() fails to enable single stepping back. The fix removes the excessive condition fromDepth >= afterPopDepth in notifyFramePop() method (stepControl.c:346) to ensure that when a method cal! led from the stepping frame (and during which we had stepping disabled) has returned the stepping is re-enabled to continue instructions steps in the original stepping frame. Webrev: http://cr.openjdk.java.net/~dtitov/8163127/webrev.01 Bug: https://bugs.openjdk.java.net/browse/JDK-8163127 Thanks! --Daniil
