Request for review: 8000797: NPG: is_pseudo_string_at() doesn't work
coleen.phillimore at oracle.com
Wed Feb 20 10:20:38 PST 2013
Since nobody replied to my code review request, I modified the change.
Summary: Add JVM_CONSTANT_PseudoString in place of JVM_CONSTANT_Object
and use this tag to distinguish patched pseudo strings. The original
string is retained if it was present.
I'm not sure how class file reconstitution for pseudo-strings is going
to work, but I thought it was prudent to leave the Symbol* in the slot
for the patched string.
I noticed that JVM_CONSTANT_Object is no longer used so reused that
constant to be PseudoString. I had an edit to go in the other
direction (if a pseudo-string is patched into a constant pool, make it a
JVM_CONSTANT_Object), but this didn't fully capture that the constant
pool slot also points to a Symbol. John or Christian, can you confirm
that you don't plan to use object constants anymore?
open webrev at http://cr.openjdk.java.net/~coleenp/8000797/
bug link at http://bugs.sun.com/view_bug.do?bug_id=8000797
Tested with vm.quick.testlist (including parallel_class_loading and mlvm
tests) and java/lang/invoke and java/lang/instrument tests.
On 2/8/2013 4:59 PM, Coleen Phillimore wrote:
> Summary: Null out the original string when patching constant pool with
> This call is_pseudo_string() conditionalized a pseudo_string_at()
> call, but since it always failed, it went to string_at() which
> returned the same thing. The only thing that seems to treat
> pseudo-string and regular strings differently is the verifier. The
> verifier pushes an object for pseudo string vs. an object of type
> open webrev at http://cr.openjdk.java.net/~coleenp/8000797/
> bug link at http://bugs.sun.com/view_bug.do?bug_id=8000797
> Tested with vm.mlvm.testlists and java/lang/invoke tests.
More information about the hotspot-runtime-dev