<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
On 2/13/14 12:00 PM, Coleen Phillimore wrote:<br>
<blockquote cite="mid:52FD1653.2030809@oracle.com" type="cite">
<meta http-equiv="content-type" content="text/html;
charset=ISO-8859-1">
<font size="+1">Summary: Remove search in system dictionary and
hacks, replace with verifying in CLD::_klasses list.<br>
<br>
open webrev at <a moz-do-not-send="true"
class="moz-txt-link-freetext"
href="http://cr.openjdk.java.net/%7Ecoleenp/8027146/">http://cr.openjdk.java.net/~coleenp/8027146/</a><br>
</font></blockquote>
<tt><font size="+1"> <br>
<font size="+1">src/share/vm/classfile/classLoaderData.<font
size="+1">h</font>pp<br>
<font size="+1"> No<font size="+1"> comments<font size="+1">
other than why add a blank line on 273<font size="+1">?<br>
<br>
</font></font></font></font>src/share/vm/classfile/classLoaderData.<font
size="+1">c</font>pp<br>
<font size="+1"> No comments.<br>
<br>
</font>src/share/vm/classfile/dictionary.cpp<br>
<font size="+1"> No comments.<br>
<br>
</font>src/share/vm/classfile/systemDictionary.<font size="+1">h</font>pp<br>
<font size="+1"> No com<font size="+1">ments.<br>
<br>
</font></font>src/share/vm/classfile/systemDictionary.<font
size="+1">c</font>pp<br>
<font size="+1"> So
SystemDictionary::verify_obj_klass_present() isn't even<br>
<font size="+1"> useful or correct in sanity
checking/debug code?<br>
<br>
<font size="+1"> Update: So InstanceKlass::verify_on()
now does a much simpler<br>
<font size="+1"> <font size="+1">contains_klass()
check. I <font size="+1">understand now why the
existing<br>
<font size="+1"> check needs to be replaced
by the simpler and less strict<br>
<font size="+1"> check.</font><br>
</font></font></font></font></font><br>
</font></font>src/share/vm/oops/arrayKlass.<font size="+1">h</font>pp<br>
<font size="+1"> No comments.<br>
<br>
</font>src/share/vm/oops/arrayKlass.<font size="+1">c</font>pp<br>
<font size="+1"> No comments.<br>
<br>
</font>src/share/vm/oops/instanceKlass.<font size="+1">h</font>pp<br>
<font size="+1"> No comments.<br>
<br>
</font>src/share/vm/oops/instanceKlass.<font size="+1">c</font>pp<br>
<font size="+1"> The old c<font size="+1">ode didn't call
verify_obj_klass_present() <font size="+1">unless<br>
<font size="+1"> is_loaded() is tr<font size="+1">ue.
The new code presumes that <font size="+1">is_loaded()<br>
<font size="+1"> is true <font size="+1">since
it expects the class to be present in the<br>
<font size="+1"> class_loader_data.<br>
<br>
<font size="+1"> Update: <font size="+1"><font
size="+1">To put it another way<font
size="+1">: The old code assume<font
size="+1">d when<br>
<font size="+1"> is_loaded() is
true that the class was in the
system dictionary<br>
<font size="+1"> or in the
placeholder table. That's not
quite correct in all<br>
<font size="+1"> cases.
When is_loaded() i<font
size="+1">s true,<font
size="+1"> all we <font
size="+1">know for
sure <font size="+1">is
that<br>
<font size="+1">
the class must be
present in<font
size="+1"> the
class_loader_data();
it might<br>
<font size="+1">
still be "on
its way" to
either the
system
dictionary or
the<br>
<font
size="+1">
placeholder
table, but has
not yet
arrived there.</font><br>
</font></font></font></font></font></font></font></font></font></font></font></font></font></font></font></font></font></font></font></font></font></font></font><br>
</font>src/share/vm/oops/klass.<font size="+1">h</font>pp<br>
<font size="+1"> No comments.<br>
<br>
</font>src/share/vm/oops/klass.<font size="+1">c</font>pp<br>
<font size="+1"> No comments.<br>
<br>
</font>src/share/vm/oops/objArrayKlass.<font size="+1">h</font>pp<br>
<font size="+1"> No comments.<br>
<br>
</font>src/share/vm/oops/objArrayKlass.<font size="+1">c</font>pp<br>
No comments.<br>
<br>
Thumbs up.<br>
<br>
<font size="+1">Dan<br>
<br>
<br>
</font><font size="+1"> </font><br>
</font></font></tt>
<blockquote cite="mid:52FD1653.2030809@oracle.com" type="cite"><font
size="+1"> bug link <a moz-do-not-send="true"
class="moz-txt-link-freetext"
href="https://bugs.openjdk.java.net/browse/JDK-8027146">https://bugs.openjdk.java.net/browse/JDK-8027146</a><br>
<br>
Tested with AllocClasses.java with VM_Verify op suggested in the
bug report (can't add it's noreg-hard). Also ran all jtreg and
gc jtreg tests with -XX:+VerifyBeforeGC.<br>
<br>
Thanks,<br>
Coleen<br>
</font> </blockquote>
<br>
</body>
</html>