Re: --enable-sjavac not enabled in tl repo yet?
On 01/18/2013 02:58 PM, Fredrik Öhrström wrote: 18 jan 2013 kl. 03:07 skrev Weijun Wang: Just tried on a latest jdk8/tl clone and The makefile listed source /space/repos/jdk8/tl/build/linux-x86_64-sjavac/langtools/gensrc/com/sun/tools/doclint/resources/doclint.java was not calculated by the smart javac Ok, as the old saying goes, this works for me :-) So could you please tar up your build directory and email it to me? I just find something interesting. /space on my machine is a symlink to /home/more/space. If I cd to /home/more/space/repos/jdk8/tl/build/linux-x86_64-sjavac and configure/make there, everything is fine. Do you still need the tar? It's already 28MB when the failure happens. Thanks Max As for the error message, since the build system has to handle building with javac as well as with sjavac, the logic for finding which java files to compile, is implemented both in make and in java (sjavac). Sjavac has the option: --compare-found-sources file_with_list_of_files which is used to make sjavac compare its own calculated list of source files with the list supplied by make. If they differ, you will get the error message was not calculated by the smart javac So why do we need to calculate the files to be compiled? Is it not just compiling the required source roots? For example like this? sjavac src/share/classes src/posix/classes src/linux/classes -d bin Lets say, that we have an opportunity to organize the source in this way. At the moment for example, when build the OpenJDK the snmp classes have to be excluded, thus the compile command looks more like: sjavac -x sun.management.snmp.* src/share/classes -d bin (In make the same calculation is handled by find and grep and sed et al.) The full filtering rules for compiling the main jdk looks like this: sjavac -x com.sun.pept.* -x com.sun.tools.example.trace.* -x com.sun.tools.example.debug.bdi.* -x com.sun.tools.example.debug.event.* -x com.sun.tools.example.debug.gui.* -x sun.dc.* -x com.sun.jmx.snmp.* -x sun.management.snmp.* -x com.sun.script.* -x com.oracle.security.* -x sun.java2d.cmm.kcms.* -xf *SolarisAclFileAttributeView.java -xf *SolarisFileStore.java -xf *SolarisFileSystem.java -xf *SolarisFileSystemProvider.java -xf *SolarisNativeDispatcher.java -xf *SolarisUserDefinedFileAttributeView.java -xf *SolarisWatchService.java -xf *SolarisAclFileAttributeView.java -xf *SolarisLoginModule.java -xf *SolarisSystem.java -xf *sun/nio/ch/DevPollArrayWrapper.java -xf *sun/nio/ch/DevPollSelectorImpl.java -xf *sun/nio/ch/DevPollSelectorProvider.java -xf *sun/nio/ch/EventPortSelectorImpl.java -xf *sun/nio/ch/EventPortSelectorProvider.java -xf *sun/nio/ch/EventPortWrapper.java -xf *sun/nio/ch/SolarisAsynchronousChannelProvider.java -xf *sun/nio/ch/SolarisEventPort.java -xf *su! n/tools/at tach/SolarisAttachProvider.java -xf *sun/tools/attach/SolarisVirtualMachine.java -xf *WrapperGenerator.java -xf *NTLoginModule.java -xf *NTSystem.java -xf *sun/nio/ch/BsdAsynchronousChannelProvider.java -xf *sun/nio/ch/KQueue.java -xf *sun/nio/ch/KQueuePort.java -xf *sun/nio/fs/BsdFileStore.java -xf *sun/nio/fs/BsdFileSystem.java -xf *sun/nio/fs/BsdFileSystemProvider.java -xf *sun/nio/fs/BsdNativeDispatcher.java -xf *sun/nio/fs/MacOSXFileSystemProvider.java -xf *sun/nio/fs/MacOSXFileSystem.java -xf *sun/nio/fs/MacOSXNativeDispatcher.java -xf *sun/tools/attach/BsdAttachProvider.java -xf *sun/tools/attach/BsdVirtualMachine.java -xf *sun/text/resources/BreakIteratorRules.java -xf *sun/text/resources/BreakIteratorRules_th.java -xf *sun/awt/AWTCharset.java -xf *sun/awt/X11/ScreenFormat.java -xf *sun/awt/X11/XArc.java -xf *sun/awt/X11/XChar2b.java -xf *sun/awt/X11/XCharStruct.java -xf *sun/awt/X11/XClassHint.java -xf *sun/awt/X11/XComposeStatus.java -xf *sun/awt/X11/XExtCodes.java! -xf *sun/ awt/X11/XFontProp.java -xf *sun/awt/X11/XFontSetExtents.java -xf *sun/awt/X11/XFontStruct.java -xf *sun/awt/X11/XGCValues.java -xf *sun/awt/X11/XHostAddress.java -xf *sun/awt/X11/XIMCallback.java -xf *sun/awt/X11/XIMHotKeyTrigger.java -xf *sun/awt/X11/XIMHotKeyTriggers.java -xf *sun/awt/X11/XIMPreeditCaretCallbackStruct.java -xf *sun/awt/X11/XIMPreeditDrawCallbackStruct.java -xf *sun/awt/X11/XIMPreeditStateNotifyCallbackStruct.java -xf *sun/awt/X11/XIMStatusDrawCallbackStruct.java -xf *sun/awt/X11/XIMStringConversionCallbackStruct.java -xf *sun/awt/X11/XIMStringConversionText.java -xf *sun/awt/X11/XIMStyles.java -xf *sun/awt/X11/XIMText.java -xf *sun/awt/X11/XIMValuesList.java -xf *sun/awt/X11/XImage.java -xf *sun/awt/X11/XKeyboardControl.java -xf *sun/awt/X11/XKeyboardState.java -xf *sun/awt/X11/XOMCharSetList.java -xf *sun/awt/X11/XOMFontInfo.java -xf *sun/awt/X11/XOMOrientation.java -xf *sun/awt/X11/XPoint.java -xf *sun/awt/X11/XRectangle.java -xf *sun/awt/X11/XSegment.ja! va -xf *su n/awt/X11/XStandardColormap.java -xf *sun/awt/X11/XTextItem.java -xf *sun/awt/X11/XTextItem16.java -xf
Re: --enable-sjavac not enabled in tl repo yet?
2013/1/18 Weijun Wang weijun.w...@oracle.com: I just find something interesting. /space on my machine is a symlink to /home/more/space. If I cd to /home/more/space/repos/jdk8/tl/build/linux-x86_64-sjavac and configure/make there, everything is fine. Ah, then sjavac is probably doing canonical or absolute on the paths, thus the makefile list will look different even though they point to the same files. That should be easy to fix! Thanks for the bug report! Do you still need the tar? It's already 28MB when the failure happens. No. //Fredrik
Re: Review Request: 8004151: build-infra: Generating X11 wrapper offset file is not cross compilable (AWT folks look here!)
Here is a new webrev with my sorts incorporated. This has been baking for quite a while in build-infra now and seems to be working. http://cr.openjdk.java.net/~erikj/8004151/webrev.jdk.04/ /Erik On 2012-12-07 11:05, Erik Joelsson wrote: (resending to full recepients list) I just hit another issue. I tried using a jdk8 boot-jdk and the order of the output was different in the generated sizes file compared to the one in the repo. Sorting both resulted in a clean diff. /Erik On 2012-12-05 14:20, Fredrik Öhrström wrote: 2012-11-29 15:54, Fredrik Öhrström skrev: 2012-11-29 15:36, Erik Joelsson skrev: I just submitted a patch to build-infra for the dual generation on all platforms since it breaks comparisons between old and new build. In general, we can't change behavior in new build without also changing the old before the old is removed. Removing sizes.64-solaris-i386 breaks the old build on solaris-x86, so I have readded it to build-infra. Thanks Erik! The new and updated webrev is here: http://cr.openjdk.java.net/~ohrstrom/webrev-8004151-gensrcX11wrapper-v2/ http://cr.openjdk.java.net/%7Eohrstrom/webrev-8004151-gensrcX11wrapper-v2/ Ok, new webrev incorporating a few more Erik changes. Any comments from the AWT experts? http://cr.openjdk.java.net/~ohrstrom/webrev-8004151-gensrcX11wrapper-v3/ //Fredrik
jdk8 - no docs directory
Hi, I built jdk8 last night and left NO_DOCS=true off but I have no docs directory. What do I need to do? Here are the highlights of what did: - install the import bundle from http://jre.us.oracle.com/java/re/jdk/8.0/promoted/latest/bundles/windows-i586/jdk-8-ea-windows-i586.tar.gz - set ALT_JDK_IMPORT_PATH pointing to it - clone top level: hg clone http://hg.openjdk.java.net/jdk8/awt - clone jdk8: hg clone http://hg.openjdk.java.net/jdk8/awt/jdk - cd ...\jdk\make - make ARCH_DATA_MODEL=32 ALLOW_DOWNLOADS=true fastdebug_build The build ran a long time and appears to have finished normally. Pete
Review Request: 8006567: jre/lib/applet missing from Mac JDK installation
When copying the j2*-image dirs to j2*-bundle, empty directories weren't included. This caused jre/lib/applet to be missing. This patch makes sure empty directories are also copied over. http://cr.openjdk.java.net/~erikj/8006567/webrev.jdk.01/ /Erik
Re: Review Request: 8003693: build-infra: bridgeBuild should allow for partial build (no hotspot)
Erik: Making it possible to use --with-import-hotspot through bridgeBuild and jprt. Also converted check for open only to use same wildcard construction as the new HOTSPOT_AVAILABLE to avoid unneeded shell expressions. http://cr.openjdk.java.net/~erikj/8003693/webrev.root.01/ Looks good- Tim
Review Request: 8006579: build-infra: In jvm.cfg, alias -server to -client when no server jvm is built.
Small patch to fix a minor difference between new and old build. In jvm.cfg, alias -server to -client when no server jvm is built. http://cr.openjdk.java.net/~erikj/8006579/webrev.jdk.01/ /Erik
Review Request: 8006583: build-infra: Remove /javax/swing/SwingBeanInfoBase.java from src.zip
Cleanup after JDK-8005096, which accidentally added /javax/swing/SwingBeanInfoBase.java to src.zip for build-infra but not for old build. http://cr.openjdk.java.net/~erikj/8006583/webrev.jdk.01/ /Erik
Re: Review Request: 8006579: build-infra: In jvm.cfg, alias -server to -client when no server jvm is built.
Erik: Small patch to fix a minor difference between new and old build. In jvm.cfg, alias -server to -client when no server jvm is built. http://cr.openjdk.java.net/~erikj/8006579/webrev.jdk.01/ Looks good. Tim
unexpected hg status report
I just ran hg status on a newly cloned jdk8 jdk directory and everything is listed as modified. I don't know if I did something wrong or something else is broken. Here's what I did: - install the import bundle from http://jre.us.oracle.com/java/re/jdk/8.0/promoted/latest/bundles/windows-i586/jdk-8-ea-windows-i586.tar.gz - set ALT_JDK_IMPORT_PATH pointing to it - clone top level: hg clone http://hg.openjdk.java.net/jdk8/awt - clone jdk8: hg clone http://hg.openjdk.java.net/jdk8/awt/jdk - cd ...\jdk\make - make ARCH_DATA_MODEL=32 ALLOW_DOWNLOADS=true fastdebug_build - make ARCH_DATA_MODEL=32 ALLOW_DOWNLOADS=true docs // found out fastdebug doesn't build docs - cd .. - hg import --no-commit ...\jdk.patch (patches for 3 files) // it failed, due to uncommitted changes - hg status // all files in jdk tree listed as modified Do I have to start over? Pete
Re: unexpected hg status report
I diffed one of the files: old mode 100644 new mode 100755 While I've occasionally run into mode differences before on files I've changed I haven't seen this problem on every file. I'm running cygwig on Win 7. Pete On 1/18/13 11:23 AM, Pete Brunet wrote: I just ran hg status on a newly cloned jdk8 jdk directory and everything is listed as modified. I don't know if I did something wrong or something else is broken. Here's what I did: - install the import bundle from http://jre.us.oracle.com/java/re/jdk/8.0/promoted/latest/bundles/windows-i586/jdk-8-ea-windows-i586.tar.gz - set ALT_JDK_IMPORT_PATH pointing to it - clone top level: hg clone http://hg.openjdk.java.net/jdk8/awt - clone jdk8: hg clone http://hg.openjdk.java.net/jdk8/awt/jdk - cd ...\jdk\make - make ARCH_DATA_MODEL=32 ALLOW_DOWNLOADS=true fastdebug_build - make ARCH_DATA_MODEL=32 ALLOW_DOWNLOADS=true docs // found out fastdebug doesn't build docs - cd .. - hg import --no-commit ...\jdk.patch (patches for 3 files) // it failed, due to uncommitted changes - hg status // all files in jdk tree listed as modified Do I have to start over? Pete
Re: Review Request: 8006583: build-infra: Remove /javax/swing/SwingBeanInfoBase.java from src.zip
Hi Erik- Cleanup after JDK-8005096, which accidentally added /javax/swing/SwingBeanInfoBase.java to src.zip for build-infra but not for old build. http://cr.openjdk.java.net/~erikj/8006583/webrev.jdk.01/ Looks good. Tim
Re: unexpected hg status report
This is probably not the right workaround but it worked for me: cd ... // one level above awt chmod -R 644 awt cd .../jdk hg status // two files found, an exe and a bat chmod 755 src/share/sample/scripting/scriptpad/src/scripts/memory.bat chmod 755 test/sun/management/windows/revokeall.exe Now hg status runs clean. cding to directories fails and would require more 755s on dirs you want to cd too but Win Explorer doesn't complain. Pete On 1/18/13 11:34 AM, Pete Brunet wrote: I diffed one of the files: old mode 100644 new mode 100755 While I've occasionally run into mode differences before on files I've changed I haven't seen this problem on every file. I'm running cygwig on Win 7. Pete On 1/18/13 11:23 AM, Pete Brunet wrote: I just ran hg status on a newly cloned jdk8 jdk directory and everything is listed as modified. I don't know if I did something wrong or something else is broken. Here's what I did: - install the import bundle from http://jre.us.oracle.com/java/re/jdk/8.0/promoted/latest/bundles/windows-i586/jdk-8-ea-windows-i586.tar.gz - set ALT_JDK_IMPORT_PATH pointing to it - clone top level: hg clone http://hg.openjdk.java.net/jdk8/awt - clone jdk8: hg clone http://hg.openjdk.java.net/jdk8/awt/jdk - cd ...\jdk\make - make ARCH_DATA_MODEL=32 ALLOW_DOWNLOADS=true fastdebug_build - make ARCH_DATA_MODEL=32 ALLOW_DOWNLOADS=true docs // found out fastdebug doesn't build docs - cd .. - hg import --no-commit ...\jdk.patch (patches for 3 files) // it failed, due to uncommitted changes - hg status // all files in jdk tree listed as modified Do I have to start over? Pete
Re: unexpected hg status report
There have been other reports of similar file permission problems when command line access via Cygwin is mixed with creating directories or doing other operations via the Windows GUI (explorer?). I don't have a reference handy, but some searching should be able to uncover the email threads in archives of build-dev or build-infra-...@openjdk.java.net My solution is to eschew the Windows GUI and use the command line, just as you would on any other server. Tim On 01/18/13 10:39, Pete Brunet wrote: This is probably not the right workaround but it worked for me: cd ... // one level above awt chmod -R 644 awt cd .../jdk hg status // two files found, an exe and a bat chmod 755 src/share/sample/scripting/scriptpad/src/scripts/memory.bat chmod 755 test/sun/management/windows/revokeall.exe Now hg status runs clean. cding to directories fails and would require more 755s on dirs you want to cd too but Win Explorer doesn't complain. Pete On 1/18/13 11:34 AM, Pete Brunet wrote: I diffed one of the files: old mode 100644 new mode 100755 While I've occasionally run into mode differences before on files I've changed I haven't seen this problem on every file. I'm running cygwig on Win 7. Pete On 1/18/13 11:23 AM, Pete Brunet wrote: I just ran hg status on a newly cloned jdk8 jdk directory and everything is listed as modified. I don't know if I did something wrong or something else is broken. Here's what I did: - install the import bundle from http://jre.us.oracle.com/java/re/jdk/8.0/promoted/latest/bundles/windows-i586/jdk-8-ea-windows-i586.tar.gz - set ALT_JDK_IMPORT_PATH pointing to it - clone top level: hg clone http://hg.openjdk.java.net/jdk8/awt - clone jdk8: hg clone http://hg.openjdk.java.net/jdk8/awt/jdk - cd ...\jdk\make - make ARCH_DATA_MODEL=32 ALLOW_DOWNLOADS=true fastdebug_build - make ARCH_DATA_MODEL=32 ALLOW_DOWNLOADS=true docs // found out fastdebug doesn't build docs - cd .. - hg import --no-commit ...\jdk.patch (patches for 3 files) // it failed, due to uncommitted changes - hg status // all files in jdk tree listed as modified Do I have to start over? Pete
Re: unexpected hg status report
Hi Tim, I did the whole process from the cygwin command line, as I normally do. I just mentioned Explorer as a way to look through directories since I 644'd everything as a workaround and so was getting access denied messages when cding to 644'd directories. This issue of hg status reporting that everything has been modified after a clone is new. hg diff indicates that the repo thinks all my files should be 644 but the clone process apparently made them all 755. At least that's what I think is going on. I don't know if the permissions are wrong in the cloned repo (.hg directory) or the cloned files in the workspace. Pete On 1/18/13 12:55 PM, Tim Bell wrote: There have been other reports of similar file permission problems when command line access via Cygwin is mixed with creating directories or doing other operations via the Windows GUI (explorer?). I don't have a reference handy, but some searching should be able to uncover the email threads in archives of build-dev or build-infra-...@openjdk.java.net My solution is to eschew the Windows GUI and use the command line, just as you would on any other server. Tim On 01/18/13 10:39, Pete Brunet wrote: This is probably not the right workaround but it worked for me: cd ... // one level above awt chmod -R 644 awt cd .../jdk hg status // two files found, an exe and a bat chmod 755 src/share/sample/scripting/scriptpad/src/scripts/memory.bat chmod 755 test/sun/management/windows/revokeall.exe Now hg status runs clean. cding to directories fails and would require more 755s on dirs you want to cd too but Win Explorer doesn't complain. Pete On 1/18/13 11:34 AM, Pete Brunet wrote: I diffed one of the files: old mode 100644 new mode 100755 While I've occasionally run into mode differences before on files I've changed I haven't seen this problem on every file. I'm running cygwig on Win 7. Pete On 1/18/13 11:23 AM, Pete Brunet wrote: I just ran hg status on a newly cloned jdk8 jdk directory and everything is listed as modified. I don't know if I did something wrong or something else is broken. Here's what I did: - install the import bundle from http://jre.us.oracle.com/java/re/jdk/8.0/promoted/latest/bundles/windows-i586/jdk-8-ea-windows-i586.tar.gz - set ALT_JDK_IMPORT_PATH pointing to it - clone top level: hg clone http://hg.openjdk.java.net/jdk8/awt - clone jdk8: hg clone http://hg.openjdk.java.net/jdk8/awt/jdk - cd ...\jdk\make - make ARCH_DATA_MODEL=32 ALLOW_DOWNLOADS=true fastdebug_build - make ARCH_DATA_MODEL=32 ALLOW_DOWNLOADS=true docs // found out fastdebug doesn't build docs - cd .. - hg import --no-commit ...\jdk.patch (patches for 3 files) // it failed, due to uncommitted changes - hg status // all files in jdk tree listed as modified Do I have to start over? Pete
Re: unexpected hg status report
This issue may have been due to using hg clone from the dos prompt instead of the cygwin prompt. I cloned again from the cygwin prompt and now hg status doesn't indicate any modified files. -Pete On 1/18/13 1:42 PM, Pete Brunet wrote: Hi Tim, I did the whole process from the cygwin command line, as I normally do. I just mentioned Explorer as a way to look through directories since I 644'd everything as a workaround and so was getting access denied messages when cding to 644'd directories. This issue of hg status reporting that everything has been modified after a clone is new. hg diff indicates that the repo thinks all my files should be 644 but the clone process apparently made them all 755. At least that's what I think is going on. I don't know if the permissions are wrong in the cloned repo (.hg directory) or the cloned files in the workspace. Pete On 1/18/13 12:55 PM, Tim Bell wrote: There have been other reports of similar file permission problems when command line access via Cygwin is mixed with creating directories or doing other operations via the Windows GUI (explorer?). I don't have a reference handy, but some searching should be able to uncover the email threads in archives of build-dev or build-infra-...@openjdk.java.net My solution is to eschew the Windows GUI and use the command line, just as you would on any other server. Tim On 01/18/13 10:39, Pete Brunet wrote: This is probably not the right workaround but it worked for me: cd ... // one level above awt chmod -R 644 awt cd .../jdk hg status // two files found, an exe and a bat chmod 755 src/share/sample/scripting/scriptpad/src/scripts/memory.bat chmod 755 test/sun/management/windows/revokeall.exe Now hg status runs clean. cding to directories fails and would require more 755s on dirs you want to cd too but Win Explorer doesn't complain. Pete On 1/18/13 11:34 AM, Pete Brunet wrote: I diffed one of the files: old mode 100644 new mode 100755 While I've occasionally run into mode differences before on files I've changed I haven't seen this problem on every file. I'm running cygwig on Win 7. Pete On 1/18/13 11:23 AM, Pete Brunet wrote: I just ran hg status on a newly cloned jdk8 jdk directory and everything is listed as modified. I don't know if I did something wrong or something else is broken. Here's what I did: - install the import bundle from http://jre.us.oracle.com/java/re/jdk/8.0/promoted/latest/bundles/windows-i586/jdk-8-ea-windows-i586.tar.gz - set ALT_JDK_IMPORT_PATH pointing to it - clone top level: hg clone http://hg.openjdk.java.net/jdk8/awt - clone jdk8: hg clone http://hg.openjdk.java.net/jdk8/awt/jdk - cd ...\jdk\make - make ARCH_DATA_MODEL=32 ALLOW_DOWNLOADS=true fastdebug_build - make ARCH_DATA_MODEL=32 ALLOW_DOWNLOADS=true docs // found out fastdebug doesn't build docs - cd .. - hg import --no-commit ...\jdk.patch (patches for 3 files) // it failed, due to uncommitted changes - hg status // all files in jdk tree listed as modified Do I have to start over? Pete