RFR(xxs): 8224795: some runtime/SelectionResolution tests are timing out

Robbin Ehn robbin.ehn at oracle.com
Mon May 27 13:32:49 UTC 2019

Hi all,

There is a fault in changeset 8221734: Deoptimize with handshakes.
Which causes us clear the method's code unconditionally.
It must be checked that it's pointing to this nmethod before clearing.
I made an error while changing some code after reviews of 8221734 here:

This patch restores correct behavior:
diff -r d871ce8ab96b src/hotspot/share/code/nmethod.cpp
--- a/src/hotspot/share/code/nmethod.cpp	Sat May 25 20:55:33 2019 +0900
+++ b/src/hotspot/share/code/nmethod.cpp	Mon May 27 12:29:26 2019 +0200
@@ -1261,7 +1261,7 @@

  void nmethod::unlink_from_method() {
    if (method() != NULL) {
-    method()->unlink_code();
+    method()->unlink_code(this);


I do not see any new issues in t1-7 and SelectionResolution (executed in t4 and 
t6) now passes. Also locally verified SelectionResolution tests with Xcomp.

Thanks, Robbin

More information about the hotspot-compiler-dev mailing list