Linux PPC JDK8 build failure

[Steve Poole]

cool - thanks Volker.

On 30 Apr 2013, at 15:04, Volker Simonis <volker.simonis at gmail.com> wrote:

> Hi Steve,
> 
> I've noticed it and I've fixed it already (will push it right now -
> the Solaris and Linux test builds just finished successfully).
> 
> I agree with you that we shouldn't change the Makefiles for this one
> occurrence of USE_PTHREADS especially because it doesn't seem to be
> used in any other place any more.
> 
> So I fixed it similarly to your suggestion with the only exception
> that instead of using "if defined(__linux__) || defined(__aix__)" I
> just inverted the logic to "if !defined(__solaris__)" which seems more
> robust in this context (would also work on HPUX for example:)
> 
> I would also advocate to run at least linux/x86_64 and Linux/ppc
> builds before submitting any changes. That's really easy now with JDK8
> and needs not more than 20 minutes:)
> 
> Regards,
> Volker
> 
> 
> 
> 
> On Tue, Apr 30, 2013 at 3:38 PM, Steve Poole <spoole at linux.vnet.ibm.com> wrote:
>> hi all - been looking at the linux build failure :
>> 
>> 
>> /home/spoole/jenkins/workspace/openjdk.ppcaix.jdk8.linux.ppc_64/jdk/src/solaris/bin/java_md_solinux.c: In function ‘ContinueInNewThread0’:
>> /home/spoole/jenkins/workspace/openjdk.ppcaix.jdk8.linux.ppc_64/jdk/src/solaris/bin/java_md_solinux.c:1031: error: ‘thread_t’ undeclared (first use in this function)
>> /home/spoole/jenkins/workspace/openjdk.ppcaix.jdk8.linux.ppc_64/jdk/src/solaris/bin/java_md_solinux.c:1031: error: (Each undeclared identifier is reported only once
>> /home/spoole/jenkins/workspace/openjdk.ppcaix.jdk8.linux.ppc_64/jdk/src/solaris/bin/java_md_solinux.c:1031: error: for each function it appears in.)
>> /home/spoole/jenkins/workspace/openjdk.ppcaix.jdk8.linux.ppc_64/jdk/src/solaris/bin/java_md_solinux.c:1031: error: expected ‘;’ before ‘tid’
>> /home/spoole/jenkins/workspace/openjdk.ppcaix.jdk8.linux.ppc_64/jdk/src/solaris/bin/java_md_solinux.c:1033: warning: implicit declaration of function ‘thr_create’
>> /home/spoole/jenkins/workspace/openjdk.ppcaix.jdk8.linux.ppc_64/jdk/src/solaris/bin/java_md_solinux.c:1033: error: ‘tid’ undeclared (first use in this function)
>> /home/spoole/jenkins/workspace/openjdk.ppcaix.jdk8.linux.ppc_64/jdk/src/solaris/bin/java_md_solinux.c:1035: warning: implicit declaration of function ‘thr_join’
>> /home/spoole/jenkins/workspace/openjdk.ppcaix.jdk8.linux.ppc_64/jdk/src/solaris/bin/java_md_solinux.c:1036: warning: cast from pointer to integer of different size
>> gmake[2]: *** [/home/spoole/jenkins/workspace/openjdk.ppcaix.jdk8.linux.ppc_64/build/linux-ppc64-normal-server-slowdebug/jdk/objs/libjli/java_md_solinux.o] Error 1
>> 
>> 
>> It seems to be  due to a difference  between our change in java_md_solinux.c  (where we have made using pthreads conditional on  USE_PTHREADS) and the autoconf process.
>> 
>> It seems that  our additions  of  -DUSE_PTHREADS on the linux and aix CFLAGS var (and others)  is not being honoured anymore.
>> 
>> In theory I'd suggest we push the "use pthreads or not" decision up to the top level platform config files.  However it seems that the only place where using pthreads is  a problem is right here within the java_md_solinux.c file.  So I think the more pragmatic answer is to just revert the change in java_md_solinux.c  back to make using pthreads conditional on  __linux__  and then add an additional __aix__ check.
>> 
>> the proposed diff looks like this below.
>> 
>> Comments please on whether we should go this way or look for a more comprehensive change..
>> 
>> 
>> 
>> --- a/src/solaris/bin/java_md_solinux.c Mon Apr 29 17:54:18 2013 +0200
>> +++ b/src/solaris/bin/java_md_solinux.c Tue Apr 30 14:36:04 2013 +0100
>> @@ -1002,7 +1002,7 @@
>> int
>> ContinueInNewThread0(int (JNICALL *continuation)(void *), jlong stack_size, void * args) {
>>     int rslt;
>> -#ifdef USE_PTHREADS
>> +#if defined(__linux__) || defined(__aix__)
>>     pthread_t tid;
>>     pthread_attr_t attr;
>>     pthread_attr_init(&attr);
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>