On 3/8/2012 7:25 PM, Alan Bateman wrote:
On 08/03/2012 08:09, Shi Jun Zhang wrote:
There is still no reply from build infra project and even if it is in
build infra, it will take a long time to merge back to trunk. But
this including pthread problem really affects AIX platform. I'm
thinking we can use #ifndef __solaris__ form because all other
POSIX-conformant platforms (BSD, Mac, AIX, ...) except Solaris need
to include pthread.h.
Here is the webrev:
http://cr.openjdk.java.net/~zhangshj/pthread/webrev.00/
Kumar will likely have an opinion on java_md.c/ContinueInNewThread0.
Personally I would go for #ifdef __solaris ... #end ... #endif rather
than #ifndef __solaris__.
I've modified it to #ifdef __solaris__ ... #else ... #endif.
I don't think the changes to NativeThread.c are quite right because it
assumes that we need to signal threads on platforms other than Solaris
and it also assumes the signal is __SIGRTMAX - 2. I don't know what
you use in your AIX port but once the preemptive close issues are
sorted out on Mac then I assume it will be SIGIO. So for this one I
would suggest the ifdef in the init, current and signal methods be
ifdef INTERRUPT_SIGNAL.
The situation in NativeThread.c is more complicated than other 2 files.
I'm not familiar with BSD or Mac. It seems that we don't need to signal
threads on BSD or Mac. And INTERRUPT_SIGNAL on AIX will definitely be
different from the one on Linux. I think we'd better separate the
changes in NativeThread.c from this patch and try to solve it later.
The change to socket_md.c looks okay to me but you will need to
re-base your patch due to the Mac port in jdk8/tl.
I'm a new comer and i got known from Charles about the difference
between jdk8 and jdk8/tl. The latest webrev is based on jdk8/tl.
http://cr.openjdk.java.net/~zhangshj/pthread/webrev.01/
--
Regards,
Shi Jun Zhang
