On 2/29/2012 12:16 AM, Artem Ananiev wrote:
I don't think so. awt.toolkit is a system property referenced in
JavaDoc (see Toolkit#getDefaultToolkit() for details), but AWT_TOOLKIT
is just an optional hint. The system property must take a precedence
over the env variable.
I agree that the AWT_TOOLKIT isn't documented properly. However, the
test has been written some time ago and relied on this behavior. And it
worked just fine before merging the Mac Port changes. So technically
this is a regression.
I don't know why the test was written this way even back in 1.6 time
frame, when both XToolkit and MToolkit were available.
Well, the test has been written by you. :)
Yes, we can just drop this behavior, in which case the fix that you're
currently reviewing doesn't make sense and should actually change the
test by eliminating the command line causing this test to fail.
The right behavior with -Dawt.toolkit=sun.awt.motif.MToolkit is to throw
a ClassNotFoundException exception, because there is not such toolkit in
JDK7. However, I'm also fine if you completely remove this scenario from
the test.
Fine by me. We'll review a new fix with you.
--
best regards,
Anthony
Thanks,
Artem
Please clarify if you want this issue to be addressed this way.
--
best regards,
Anthony
be checked. Please note that the test in question fails on X11 exactly
for this reason, it invokes:
AWT_TOOLKIT=XToolkit ${TESTJAVA}/bin/java -Djava.awt.headless=true \
-Dawt.toolkit=sun.awt.motif.MToolkit \
TestWrapped sun.awt.X11.XToolkit
and expects the AWT_TOOLKIT to have precedence over the -Dawt.toolkit
setting. So we must check the environment variable on any *NIX
platform.
2. On Mac, CToolkit is not explicitly set in awt_LoadLibrary.c - I
realize it's used by default (inherited from java_props_md.c), but it
would be fine to leave a comment in the code.
The comment at line 112 has been added for exactly this reason. It's
already there.
3. AFAIK, there is no need to check global refs for NULL:
DeleteGlobalRef() is ready to handle null refs.
It may be implemented this way. However, its specification [1] doesn't
mention that NULL is a valid argument. Therefore, I prefer to check for
NULL before calling the function.
[1]
http://download.java.net/jdk8/docs/technotes/guides/jni/spec/functions.html#DeleteGlobalRef
OK, let it be so.
Thanks,
Artem
--
best regards,
Anthony
Thanks,
Artem
--
best regards,
Anthony
On 2/28/2012 2:01 AM, Artem Ananiev wrote:
Hi, Anthony,
I have an impression your webrev is prepared against an outdated
version of code. For example, the change in java_props_md.c:431 is
already in the workspace...
Other comments:
1. HeadlessGraphicsEnvironment is in the sun.java2d package, not
sun.awt
2. Changes to awt_LoadLibrary.c look fine as all the defaults are
set
in java_props.c
3. GraphicsEnvironment.java also looks fine.
Thanks,
Artem
On 2/27/2012 6:30 AM, Anthony Petrov wrote:
Hello,
Please review a fix for
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=7147435 at:
http://cr.openjdk.java.net/~anthony/7u4-2-headlessTestFailed-7147435.0/
This bug is a regression of 7124511 fixed for the JDK Mac Port [1].
With
that fix the code setting the awt.toolkit and java.awt.graphicsenv
system properties has been removed from JNI_OnLoad() of libawt.so.
Actually, the test WrappedToolkitTest.sh relies on the ability to
override the default toolkit by means of setting the AWT_TOOLKIT
environment variable, and because of the removal the test has
failed.
With the fix for 7147435 I'm restoring the removed parts. In
order to
not break the fix for 7124511, I'm setting the system properties
only if
the XToolkit has been requested explicitly.
[1] http://cr.openjdk.java.net/~anthony/x-5-forceHeadless.0/
--
best regards,
Anthony
