On Wed, Nov 13, 2013 at 12:52 PM, Andrew Haley <a...@redhat.com> wrote: > On 11/13/2013 11:29 AM, Richard Biener wrote: >> On Wed, Nov 13, 2013 at 12:24 PM, Andrew Haley <a...@redhat.com> wrote: >>> On 11/13/2013 10:56 AM, Richard Biener wrote: >>>> At least we don't need a Java source code frontend, no? Just keeping >>>> the bytecode compiler and GIJ should be enough? That way we can >>>> strip the classpath copy of everything that isn't needed, thus not >>>> provide a Java library. Reduces testing coverage of GCJ to almost >>>> zero, but ... >>> >>> Eh? We don't even have a Java source code frontend. In a GCC >>> build we compile everything from bytecode. >> >> Don't we drop in ecj.jar and compile that to native code? Ah, seems to >> be an optional feature. Which means only very little pieces of libgcj should >> be required to bootstrap if we remove that feature without also dropping >> in a classpath.jar? > > I don't get it. If you want not to build libgcj in bootstrap, don't > build it. But there's no need to mess about like this.
I also want to reduce repository size by removing parts of (or you say all of?) classpath, retaining only those portions we need for bootstrap & regtest. If people want the ecj frontend they then not only have to supply the ecj.jar but also a classpath.jar. Now checking whether we can run the java testsuite without libjava (there doesn't seem to be a way to just disable building the classpath piece). Richard. > Andrew. > >
