On Mon, 2018-07-23 at 18:27 +0200, Severin Gehwolf wrote: > Hi, > > Could I please get a review of this one-liner change related to jhsdb > --mixed when attaching to a running Java process? The issue arises when > threads are in native code and that native code has frame pointers not > properly preserved. In such a case the SA performs a simple frame > pointer valididy check: ebp >= esp > > However, the code of retrieving the value for esp is incorrect in as > much as it's not in sync with native code in regards to the register > index: > > native code => X86ThreadContext.SP > Java code => X86ThreadContext.ESP > > X86ThreadContext.ESP is never being set by the native code. Since > X86ThreadContext.getRegisterAsAddress(X86ThreadContext.ESP) then > returns null, ebp.lessThan(esp) wrongly returns false causing the > issue. This webrev fixes it by using SP as index on the Java side. > Thoughts? > > webrev: http://cr.openjdk.java.net/~sgehwolf/webrevs/JDK-8208091/webrev.01/ > bug: https://bugs.openjdk.java.net/browse/JDK-8208091
Anyone willing to review this one-liner? Thanks, Severin > Thanks, > Severin
