Am 09.07.2017 um 18:07 schrieb Damjan Jovanovic: > On Sun, Jul 9, 2017 at 5:44 PM, Matthias Seidel <[email protected]> > wrote: > >> Hi Damjan, >> >> I can only comment on a non-technical POV... >> >> Am 09.07.2017 um 14:53 schrieb Damjan Jovanovic: >>> Hi >>> >>> Currently our Java policy [1] mandates a baseline of Java 5, which was >>> released in 2004, and was EOL from 2008; even paid Oracle customers >> stopped >>> getting updates in 2015 [2]. >> That page was last updated in 2008, so it might definitely need an >> overhaul. ;-) >> >>> There are many language and class library improvements in later versions >> of >>> Java that we would benefit from in AOO, and if there are no objections, I >>> would like to increase the minimum version to 7, if not 8. >> At least the OS/2 port still relies on Java 6 >> [http://trac.netlabs.org/java]. >> But I think the support is still at beta stage... >> >> > What is the status of our OS/2 support?
We should ask Yuri Dario for that... ;-) Support for Java is still experimental I think. > > The language features are the most important, and it should be possible to > use Java 7 language features while only using Java <= 5 APIs, and build > into Java 5 compatible class files, as long as javac (ie. the JDK) is from > Java >= 7. > > But even if it isn't, since class files are the same on all platforms, it > should be possible to jerryrig a fake "javac" on OS/2, that compiles the > java file on a remote javac 7, and then downloads the Java 5 compatible > class file. > > >>> Java 7 would give us the badly needed try-with-resources feature, better >>> type inference for generics, and Java 8: lambdas and method references, >>> better java.util collections and lambda-based APIs for them, unsigned >>> integer arithmetic, etc. >>> >>> The only possible downside I see is that the few alternative JVMs that do >>> exist, tend to only work with older version of Java, for example, the >>> Apache Harmony class library used in earlier versions of Android only >>> supported Java 5, GNU Classpath also 5, CacaoVM only supports 6 [3]. >>> Alternative Java implementation aren't relevant to desktop users >>> (x86/amd64/sparc) any more, since OpenJDK provides free/open-source JVMs >>> for those, but other platforms and more exotic use cases do need them, >> eg. >>> the ARM JVM is only available from Oracle as a commercial product, and >> the >>> BugVM project that compiles Java to native code for the iPhone, uses the >>> Harmony-based Android library. >>> >>> Having said that, there are ways to compile most Java 7 code into Java 5 >>> classes. For example try-with-resources compiles into a normal >>> try-catch-finally, with calls to the Exception.addSuppressed() API that >>> only exists on Java 7, but the class file can be changed to eliminate >> them; >>> I already have a tool that does this. So I guess it's only 8 that's >>> controversial, as it's lambdas compile into "invokevirtual" instructions, >>> which didn't exist prior to Java 7. >>> >>> Anyway I am proceeding with (uncommitted) development under the >> assumption >>> at least 7 is ok, and will revisit with 8 if possible. Please object soon >>> if you do :-). >> I still have problems to build AOO for Windows with Java 8. >> One user mentioned a problem with JavaDoc due to an API change in Java 8. >> [http://blog.joda.org/2014/02/turning-off-doclint-in-jdk-8-javadoc.html] >> >> > You must be building the 4.1 branches. Trunk has been building perfectly > with Java 8 for years, since I patched the Javadoc issues we had. Yes, I am building 4.1 branch... But with trunk I get other errors regarding gtest... ;-) Would it be possible to backport your patches? Otherwise I will continue to build 4.1.4 with Java 7. Regards, Matthias > > >> Regards, Matthias >> >>> Damjan >>> >>> References: >>> [1] https://wiki.openoffice.org/wiki/Policies/Java_Usage >>> [2] https://en.wikipedia.org/wiki/Java_version_history >>> [3] https://en.wikipedia.org/wiki/Comparison_of_Java_virtual_machines >>> >> >>
smime.p7s
Description: S/MIME Cryptographic Signature
