<html>
<head>
<meta content="text/html; charset=windows-1252"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<div class="moz-cite-prefix">Hi Eric,<br>
<br>
This is a good example to learn how internal VM tests work, so
thanks!<br>
<br>
On 2/12/16 5:37 AM, Erik Helin wrote:<br>
</div>
<blockquote cite="mid:20160212103721.GB32525@ehelin.jrpg.bea.com"
type="cite">
<pre wrap="">Hi all,
this small patch adds a couple of unit tests for G1MMUTracker.
Webrev:
<a class="moz-txt-link-freetext" href="http://cr.openjdk.java.net/~ehelin/8149128/00/">http://cr.openjdk.java.net/~ehelin/8149128/00/</a>
Testing:
- JPRT
- Running `make test-hotspot-internal` locally
Thanks,
Erik
</pre>
</blockquote>
Some quick comments: <br>
<meta http-equiv="content-type" content="text/html;
charset=windows-1252">
src/share/vm/gc/g1/g1MMUTracker.cpp:<br>
- Should <span class="new">assert_jlong_within(), </span><span
class="new">assert_jlong_eq(), and </span><span class="new">assert_double_eq()
move to debug.hpp? They seem useful.<br>
- Is starting from milliseconds and converting to seconds buying
anything? <br>
</span>
<ul>
<li><span class="new">const double time_window = ms(500).to_sec();</span></li>
<li><span class="new">const double time_window = 0.5;</span></li>
</ul>
<span class="new"> - From my point of view the 2nd is more
clear. I guess the asserts would have to use doubles, and perhaps
allow a little slop, but that moves the boilerplate code out of
the test, which should make the test's intentions much more
obvious.<br>
<br>
- Derek<br>
</span>
<meta http-equiv="content-type" content="text/html;
charset=windows-1252">
</body>
</html>