AIX Changes

[Steve Poole]

Update:   

Unfortunately   I've quickly hit two java compilation problems with the big patch so far.

 One is a missing brace in src/share/classes/java/security/AccessControlContext.java.  

The other is , probably , to do with missing changes in  src/share/classes/sun/net/www/http/HttpClient.java that gives me the following
../../../src/share/classes/sun/net/www/protocol/http/HttpURLConnection.java:1772: error: failedOnce is not public in HttpClient; cannot be accessed from outside package
		    http.failedOnce = this.failedOnce;

Neither of the changes that caused these problems seem to be related to the basics of getting an AIX build going. So I'm going to start again.  This time I  will take the IBM changes and feed them in one by one and use the big patch as a reference to see what SAP did in the same circumstances.    Hopefully I can get a basic AIX build running fairly quickly.


On 2 Jul 2012, at 13:08, Steve Poole wrote:

> I'm doing a full build too - so looks like you're actually hitting similar problems.
> 
> I'm going to get the class libs building first   (setting BUILD_HOTSPOT=false)  and let Volker sort out how to get hotspot building as part a full build!
> 
> 
> On 2 Jul 2012, at 11:08, Lindenmaier, Goetz wrote:
> 
>> Hi Steve, 
>> 
>> tiered.make is a makefile needed to build the jit compiler.  As you set all_debugcore and 
>> CORE_BUILD=true, the jit compiler should not be built.  So I assume, these flags did not
>> make their way to the hotspot build.
>> Have a look at the build log Volker published:
>> http://cr.openjdk.java.net/~simonis/ppc-aix-port/build-logs/output_ppc-aix-port_dbg.log
>> 
>> It says "Entering hotspot for target(s) all_debugcore" at some point.  Can you see that in 
>> your build log? Else the flag is not properly passed to the hotspot build.
>> 
>> I am building hotspot on aix by going to .../hotspot/make and calling 
>> make ALT_BOOTDIR=/sapmnt/depot/tools/gen/rs6000_64/licenseware/jse/1.6.0 ALT_OUTPUTDIR=/usr/work/d045726/oJ/builds_aix-hotspot/build-is3036-asm ARCH_DATA_MODEL=64 HOTSPOT_BUILD_JOBS=8  VERBOSE=true CC_INTERP=true OPENJDK=true CORE_BUILD=true all_debugcore
>> so I'm sure that is working.
>> 
>> Unfortunately, we can't easily make the full build on aix ... Volker is trying.
>> Maybe you can mail me a build log?  
>> I'll also have a further look at the makefiles.
>> 
>> Cheers,
>> Goetz.
>> 
>> 
>> -----Original Message-----
>> From: Steve Poole [mailto:spoole at linux.vnet.ibm.com] 
>> Sent: Montag, 2. Juli 2012 10:23
>> To: Lindenmaier, Goetz
>> Cc: ppc-aix-port-dev at openjdk.java.net
>> Subject: Re: AIX Changes
>> 
>> 
>> 
>> On 2 Jul 2012, at 09:08, Lindenmaier, Goetz wrote:
>> 
>>> Hi Steve,
>>> 
>>> sorry, I can't help you with that.  Freetype is only used in OpenJDK, therefore
>>> we did not port it.  We use code that is not available in the open, so we may
>>> not share it.
>>> 
>> Ok  that's fine.  Next question..
>> 
>> How do I build Hotspot on AIX?    I used the instructions in the PPC guide in the root of the repo - specifically 
>> 
>> 
>> HOTSPOT_TARGET=all_debugcore 
>> CC_INTERP=true 
>> OPENJDK=true 
>> CORE_BUILD=true 
>> 
>> but I'm getting a build failure of:
>> 
>> make[6]: *** No rule to make target `/home/spoole/hudson/workspace/sp.ppcaix.jdk7u.aix.ppc64/work/hotspot/make/aix/makefiles/tiered.make.o', needed by `/home/spoole/hudson/workspace/sp.ppcaix.jdk7u.aix.ppc64/work/hotspot/make/aix/makefiles/tiered.make'.  Stop.
>> 
>> Does that problem sound familiar?
>> 
>> 
>>> Sorry for that,
>>> Goetz.
>>> 
>>> -----Original Message-----
>>> From: ppc-aix-port-dev-bounces at openjdk.java.net [mailto:ppc-aix-port-dev-bounces at openjdk.java.net] On Behalf Of Steve Poole
>>> Sent: Samstag, 30. Juni 2012 10:47
>>> To: ppc-aix-port-dev at openjdk.java.net
>>> Subject: Re: AIX Changes
>>> 
>>> Hi Goetz,
>>> 
>>> As we discussed via the phone yesterday I started to do a sanity build with this patch applied.    
>>> 
>>> One thing I've found already is that the freetypecheck fails to build as its makefile is not AIX aware and tries to use  -rpath. 
>>> Your patch correctly fixes the same sort of problem in Program.gmk   so I assume that either you missed  make/tools/freetypecheck/Makefile  or you have another sort of work-around?
>>> 
>>> 
>>> 
>>> 
>>> On 28 Jun 2012, at 11:00, Steve Poole wrote:
>>> 
>>>> 
>>>> Hi guys -   we've taken a look at Goetz's AIX changes and we would really like them to be  committed.   I think it would be best if someone for SAP did that :-)
>>>> 
>>>> Attempting to reconcile the different approaches between the IBM and the SAP code bases in a piecemeal manner just means it will take a long time before we have a working codebase - which is somewhat mad given you already have one.
>>>> 
>>>> If you're happy with doing a bulk commit ,  then, once the changes are in the codebase, we can work through any remaining IBM changes we think are needed and offer them up.   
>>>> 
>>>> What do you think?
>>>> 
>>>> 
>>>> Cheers
>>>> 
>>>> Steve
>>> 
>> 
>