For users of --with-openjdk

[DJ Lucas]

On 08/11/2010 01:11 PM, Dr Andrew John Hughes wrote:

>> Could you elaborate on what you mean by 'conflicting types'?

Bad memory?  (human, not system) :-)


> I didn't try the bootstrap-with-icedtea option myself so far, just the
> existing options (bootstrap-with-gcj, no-bootstrap-with-icedtea) to
> make sure we didn't regress.  I left out the tools.jar chunk of the
> patch because I wasn't sure about it.
> 
> There's a bit of history to the whole tools.jar situation.  The ecj
> build originally created a tools.jar as well (in a similar way to the
> current rt.stamp target and friends) but it resulted in a lot of
> additional confusion and duplication of compilation with little
> benefit.  So the bootstrap was changed to pack all the bootstrap
> classes in an updated rt.jar from the system jdk.  tools.jar only
> actually needs to exist for an existence check by HotSpot.  IcedTea6
> currently satifies that by a symlink to rt.jar, as you've seen, but I
> think I'll switch it to patching HotSpot during bootstrap as I've done
> in IcedTea7.
> 
> This works fine for the existing configurations but breaks with
> bootstrapping with IcedTea.  Why? Well...
> 
> 1.  The necessary sun.* classes are built in the rt.stamp stage.  However,
> this only happens if sun.awt.Toolkit is not found (which is used to signify
> that the bootstrap JDK has Sun classes).  See
>> checking if sun.awt.SunToolkit is missing... no
> in your output.
> 
> 2.  Of course those classes are available in your JDK but in tools.jar not
> rt.jar.  The javac invocation looks at tools.jar but because we don't symlink
> the whole system JDK to bootstrap/jdk1.6.0 (as in the second stage), it's
> not accessible from the build.
> 
> Thus, your fix seems appropriate and I'll add the symlink to tools.jar
> if it exists.  I'd be very interested to know what issues you had with
> doing this.
> 

Actually, it breaks while building awt:

# Running javac:
/home/dj/icedtea6-cb463b94b82d/bootstrap/jdk1.6.0/bin/javac
-J-XX:ThreadStackSize=768 -J-client -J-Xmx896m -J-Xms128m
-J-XX:PermSize=32m -J-XX:MaxPermSize=160m -g -source 1.5 -target 5
-encoding ascii
-Xbootclasspath:/home/dj/icedtea6-cb463b94b82d/openjdk.build-ecj/classes
-sourcepath
/home/dj/icedtea6-cb463b94b82d/generated:../../../src/solaris/classes:../../../src/share/classes
-bootclasspath
/home/dj/icedtea6-cb463b94b82d/bootstrap/jdk1.6.0/jre/lib/rt.jar::/home/dj/icedtea6-cb463b94b82d/openjdk.build-ecj/classes
-d /home/dj/icedtea6-cb463b94b82d/openjdk.build-ecj/classes
@/home/dj/icedtea6-cb463b94b82d/openjdk.build-ecj/tmp/sun/java.awt/.classes.list
----------
1. ERROR in ../../../src/share/classes/java/awt/Robot.java (at line 359)
	SunWritableRaster.makeTrackable(buffer);
	                  ^^^^^^^^^^^^^
The method makeTrackable(DataBufferInt) is undefined for the type
SunWritableRaster
----------
----------
2. ERROR in ../../../src/share/classes/java/awt/Component.java (at line
3509)
	if (!caps.isPageFlipping() && sge.isFlipStrategyPreferred(peer)) {
	                                  ^^^^^^^^^^^^^^^^^^^^^^^
The method isFlipStrategyPreferred(ComponentPeer) is undefined for the
type SunGraphicsEnvironment
----------
3. ERROR in ../../../src/share/classes/java/awt/Component.java (at line
4147)
	private class FlipSubRegionBufferStrategy extends FlipBufferStrategy
	              ^^^^^^^^^^^^^^^^^^^^^^^^^^^
The type Component.FlipSubRegionBufferStrategy must implement the
inherited abstract method SubRegionShowable.validateAndShow(int, int,
int, int)
----------
4. ERROR in ../../../src/share/classes/java/awt/Component.java (at line
4178)
	private class BltSubRegionBufferStrategy extends BltBufferStrategy
	              ^^^^^^^^^^^^^^^^^^^^^^^^^^
The type Component.BltSubRegionBufferStrategy must implement the
inherited abstract method SubRegionShowable.validateAndShow(int, int,
int, int)
----------
4 problems (4 errors)make[5]: *** [.compile.classlist] Error 255
make[5]: Leaving directory
`/home/dj/icedtea6-cb463b94b82d/openjdk-ecj/jdk/make/java/awt'
make[4]: *** [all] Error 1
make[4]: Leaving directory
`/home/dj/icedtea6-cb463b94b82d/openjdk-ecj/jdk/make/java'
make[3]: *** [all] Error 1
make[3]: Leaving directory
`/home/dj/icedtea6-cb463b94b82d/openjdk-ecj/jdk/make'
make[2]: *** [jdk-build] Error 2
make[2]: Leaving directory `/home/dj/icedtea6-cb463b94b82d/openjdk-ecj'
make[1]: *** [build_product_image] Error 2
make[1]: Leaving directory `/home/dj/icedtea6-cb463b94b82d/openjdk-ecj'
make: *** [stamps/icedtea-ecj.stamp] Error 2
dj [ ~/icedtea6-cb463b94b82d ]$

I'm not sure if this was the same error I had seen earlier, but I've
since scrapped that diff in favor of the one you checked in the other
day.  This error messages below were achieved by simply copying
/opt/icedtea6-1.8.1-bin/lib/tools.jar into the previously failed build's
bootstrap/jdk-1.6.0/lib directory.  Options, from my admittedly limited
POV are to either increase the complexity of that test, or scrap it and
always build tools into the temp rt.jar (as you hinted at above).  I
like the second option as on a fairly modern system, this would probably
only amount to a few minutes extra time, and further isolate the host
JDK.  Compared to the 8-10 hours on old hardware back in the 1.6.0 SCSL
days, my (est.) one hour bootstrap is nothing.  My fear for the first
option is that the test would have to be adjusted with every change that
is not completely backwards compatible from here on out (and the above
errors only cover 6-1.8.1, no idea about other Oracle/Sun JDKs, open or
otherwise).  Of course, there is always the possibility that my System
JDK is broken and this is an isolated incident.

I'm also getting an error in swing/plaf, with the changes and the
sun.awt.SunTookit check gone, but I should probably update to the latest
snapshot before going any further (this is in 1.9 branch at the branch
point).

1. ERROR in
../../../../src/share/classes/javax/swing/plaf/synth/SynthLookAndFeel.java
(at line 245)
	((NimbusLookAndFeel) laf).shouldUpdateStyleOnEvent(event)) {
	                          ^^^^^^^^^^^^^^^^^^^^^^^^
The method shouldUpdateStyleOnEvent(PropertyChangeEvent) is undefined
for the type NimbusLookAndFeel

-- DJ Lucas

-- 
This message has been scanned for viruses and
dangerous content, and is believed to be clean.