<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body>
<div dir="ltr" style="">
<div style=""></div>
<div style="">
<div>
<div style="" dir="ltr">Hi Dan, thanks for looking <span>into this. The comment is definitely wrong about MethodTypes. In fact in the premain branch, CDS caches all the Method<span>Types::internTable<span id="ms-outlook-ios-cursor"></span> in the because they
are tested for equality. </span></span></div>
</div>
<div style="" dir="ltr"><br>
</div>
<div style="" dir="ltr"><span><span>The comment should probably be changed to something like:<span></span></span></span></div>
<div style="" dir="ltr"><span><span><br>
</span></span></div>
<div style="" dir="ltr"><span><span>MethodTypes does not have a public constructor. All MethodTypes are created by factory methods and are stored in MethodType::internTable. Therefore, any MethodTypes in static fields can be safely archived as long as internTable
is archived.<span></span></span></span></div>
<div style="" dir="ltr"><span><span><br>
</span></span></div>
<div style="" dir="ltr"><span><span>Ioi </span></span></div>
</div>
<div id="mail-editor-reference-message-container" class="ms-outlook-mobile-reference-message" style="">
<hr style="display:inline-block;width:98%" tabindex="-1">
<div id="divRplyFwdMsg" dir="ltr"><font face="Calibri, sans-serif"><b>From:</b> leyden-dev <leyden-dev-retn@openjdk.org> on behalf of Dan Heidinga <heidinga@redhat.com><br>
<b>Sent:</b> Tuesday, August 8, 2023 1:17 PM<br>
<b>To:</b> leyden-dev@openjdk.org <leyden-dev@openjdk.org><br>
<b>Subject:</b> Re: git: openjdk/leyden: created branch premain based on the branch computed-constants containing 1 unique commit
<div> </div>
</font></div>
<div dir="ltr">I'm reading through the code in the premain branch (which looks very cool though I'm still wrapping my head around it) and had an early question about the cdsHeapVerifier.cpp:
<div><br>
</div>
<div>> void CDSHeapVerifier::add_static_obj_field(InstanceKlass* ik, oop field, Symbol* name) {<br>
> if (field->klass() == vmClasses::MethodType_klass() ||<br>
> field->klass() == vmClasses::LambdaForm_klass()) {<br>
> // LambdaForm and MethodType are non-modifiable and are not tested for object equality, so<br>
> // it's OK if the static fields are reinitialized at runtime with alternative instances.<br>
> // (TODO: double check is this is correct)<br>
> return;<br>
> }<br>
</div>
<div><br>
</div>
<div>I can't speak to LambdaForms but MethodTypes definitely rely on equality for correct behaviour when invoking - see Invokers.java::checkExactType which is part of every MethodHandles.invokeExact call. Is this kind of equality test something this code should
be concerned about? </div>
<div><br>
</div>
<div>I may be looking at this in the wrong way but thought I'd chime in as I read the diff in case this helps.</div>
<div><br>
</div>
<div>--Dan</div>
</div>
<br>
<div class="gmail_quote">
<div dir="ltr" class="gmail_attr">On Tue, Aug 8, 2023 at 2:30 PM duke <<a href="mailto:duke@openjdk.org" style="">duke@openjdk.org</a>> wrote:<br>
</div>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex; border-left:1px solid rgb(204,204,204); padding-left:1ex">
The following commits are unique to the premain branch:<br>
========================================================<br>
626b62aa: Initial commit<br>
<br>
</blockquote>
</div>
</div>
</div>
</body>
</html>