- use system libz instead of internal one. internal libz conflicts with
system libz, for example apps using JNI & libz will fail (OOo).
- enable parallel hotspot compiles on mp systems (one per cpu). no pthreads
involved so its safe to do here.
- socket option corrections. may help Active MQ work.
- increase default thread stack size on amd64
- filter out corrupt env vars with no name
- Properly deal with error cases in transferTo0(). Corrects non-blocking
IO case.
All but first one from FreeBSD
since KOI8_U doesn't come with the Solaris class files, make sure it is
built in the recompile stage. Reported by Matt Jibson.
- add a useable cacerts file and use that instead of current empty one.
charset portions tested by and okay ian@.
obviating need to build with JDK Linux compat (which was i386 only anyway).
Add native_bootstrap pseudo-flavor so you can build with a native Sun JDK if you
have one (tested by building this way using JDK built with default Kaffe-based build).
Work by kurt@, tested by him on i386 and me on amd64. OK kurt@.
solaris class files to bootstrap the jdk build. this avoids the need for
devel/jdk/1.4 to bootstrap and will provide a way to bootstrap the jdk on
amd64 and sparc64 when kaffe runs there.
- fix a keyboard entry problem when numlock is on
- fixup socket option code similar to FreeBSD & NetBSD
while creating/destroying threads.
- add missing pthread_attr_destroy while creating threads. closes a
small memory leak.
Memory leaking problems reported by Carsten Raskgaard and also
David SH Rosenthal from the LOCKSS project.
undiscovered bug in the SignatureHandlerLibrary. Code generated at runtime
was put into malloc()'ed memory and later attempted to execute it without
first making it executable with mprotect(). Debugging suggestions and help
from tom@, weingart@, marc@, millert@. Thx!
- make sure linux emulation is on for the build phase. needed if you reboot
after extracting.
okay tom@
stack. When forking, the child process deallocates the threads stacks it
inheritated from the parent. The junk malloc option then attempts to write
to the deallocated memory and segfaults because its not all writable. The
child proc stays in a SIGSEGV loop when this happens. okay ian@
Demos all work but deadlocking the vm under load has been
noted. Both MToolkit and XToolkit work ok so far.
- fix segfault with MToolkit where processTree was called
with to==NULL.
- fix NPE in plugin paintGrayBox where g==null
- fix javaws startup (from FreeBSD)
- forgot to mention in last commit: Forte Peformance Tools
collector is no longer disabled.
- link awt_robot with -pthread
- disable a runtime Motif version check that incorrectly thought we were
using Motif 1.2 based on Xm major version. problem reported by Laurence
Tratt
- plugin and webstart comming later
as this is the only port, where we have a versioned distfile in
systrace.policy AND this jdk is EOL, there is no need for a generic
solution
discussed with and ok kurt@
the FULLPKGPATH, thus providing changes to packing-lists which shouldn't
happen, and making update more difficult.
Accordingly, bump all pkgnames with PSEUDO_FLAVORS, and provide an
update @pkgpath for the bug for most of them (left out the ones with 3
or 4 pseudo flavors for space constraints...)
- Based on the MODJAVA_VER, MODJAVA_JRERUN, NO_BUILD
and MACHINE_ARCH, the following things will be setup:
ONLY_FOR_ARCHS if not already set.
BUILD_DEPENDS on a jdk (native preferred).
JAVA_HOME to pass on to the port build.
RUN_DEPENDS for all jdk's and jre's that can run the port.
Review and feedback nikolay@, ian@.
