RFR (M): 8036767 PPC64: Support for little endian execution model

[Volker Simonis]

On Wed, Mar 26, 2014 at 8:07 AM, Magnus Ihse Bursie
<magnus.ihse.bursie at oracle.com> wrote:
> I have not seen it before, but that webrev contains a modification of autoconf-config.guess, which is not okay. That file is copied directly from the autoconf project and should not be modified.
>
> If that script is not working properly, a correction can be made in the config.guess wrapper. See existing code there for how it can be done.
>

We could do that, but I think that would be an ugly hack. I'm not sure
what the current autoconf-config.guess actually returns for
ppc64le-linux, but from a quick hack (i.e. changing uname to return
'ppc64le' if given the '-m' option) I can see that it returns an emtpy
string and writes an error text to stderr:

==========================================================

common/autoconf/build-aux/config.guess: unable to guess system type

This script, last modified 2012-02-10, has failed to recognize
the operating system you are using. It is advised that you
download the most up to date version of the config scripts from

  http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.guess;hb=HEAD
and
  http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.sub;hb=HEAD

If the version you run (common/autoconf/build-aux/config.guess) is
already up to date, please
send the following data and any information you think might be
pertinent to <config-patches at gnu.org> in order to provide the needed
information to handle your system.

config.guess timestamp = 2012-02-10

uname -m = ppc64le
uname -r = 3.2.0-60-generic
uname -s = Linux
uname -v = #91-Ubuntu SMP Wed Feb 19 03:54:44 UTC 2014

/usr/bin/uname -p =
/bin/uname -X     =

hostinfo               =
/bin/universe          =
/usr/bin/arch -k       =
/bin/arch              =
/usr/bin/oslevel       =
/usr/convex/getsysinfo =

UNAME_MACHINE = ppc64le
UNAME_RELEASE = 3.2.0-60-generic
UNAME_SYSTEM  = Linux
UNAME_VERSION = #91-Ubuntu SMP Wed Feb 19 03:54:44 UTC 2014

==================================================================

We could of course handle that, but I think it would be a bad idea to
treat every unrecognized system as 'ppc64le'.

I've checked that the latest config.guess (from
http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.guess;hb=HEAD)
does indeed recognize ppc64le.

So I'd suggest we just check in this new version for the current
common/autoconf/build-aux/config.guess.

It is by far the cleanest solution and it shouldn't do any harm for
the few platforms we support (and it additionally can detect mingw-64
which may be interesting for our MinGW experiments). What do you
think?

Regards,
Volker

> /Magnus
>
>> On 26 mar 2014, at 01:00, Vladimir Kozlov <vladimir.kozlov at oracle.com> wrote:
>>
>> I mean build changes in top directory:
>>
>> http://cr.openjdk.java.net/~martin/asmundak/8036767/webrev.01/
>>
>> Thanks,
>> Vladimir
>>
>>> On 3/25/14 4:18 PM, Vladimir Kozlov wrote:
>>> I assume jdk changes are reviewed by Magnus. Please, confirm.
>>>
>>> Changes require a regeneration of generated-configure.sh files.
>>>
>>> I am going to push jdk(top) changes into jdk9/hs-comp and will ask
>>> gatekeepers to push it up later.
>>>
>>> JDK control build testing is running in JPRT now.
>>>
>>> Thanks,
>>> Vladimir
>>>
>>>> On 3/25/14 12:41 PM, Vladimir Kozlov wrote:
>>>> I will push changes from webrev.05 into hs-comp/hotspot
>>>>
>>>> Where I should push jdk changes (jdk9/dev/ or hs-comp)?:
>>>>
>>>> http://cr.openjdk.java.net/~martin/asmundak/8036767/webrev.01/
>>>>
>>>> Thanks,
>>>> Vladimir
>>>>
>>>>> On 3/25/14 3:32 AM, David Holmes wrote:
>>>>> Thanks. Build side looks fine to me.
>>>>>
>>>>> David
>>>>>
>>>>>> On 25/03/2014 6:09 AM, Alexander Smundak wrote:
>>>>>> Done. Uploaded
>>>>>> http://cr.openjdk.java.net/~martin/asmundak/8036767/hotspot/webrev.05
>>>>>>
>>>>>> On Sun, Mar 23, 2014 at 8:37 PM, David Holmes
>>>>>> <david.holmes at oracle.com> wrote:
>>>>>>> On 22/03/2014 10:47 AM, Alexander Smundak wrote:
>>>>>>>>
>>>>>>>> On Fri, Mar 21, 2014 at 2:55 PM, Lindenmaier, Goetz
>>>>>>>> <goetz.lindenmaier at sap.com> wrote:
>>>>>>>>>
>>>>>>>>> My build ran into the 'little' case below, as I understand because
>>>>>>>>> ' big'
>>>>>>>>> was compared
>>>>>>>>> with 'big'.  Make is strange ...
>>>>>>>>
>>>>>>>>
>>>>>>>> Removed the spaces, please take a look at
>>>>>>>> http://cr.openjdk.java.net/~martin/asmundak/8036767/hotspot/webrev.04
>>>>>>>
>>>>>>>
>>>>>>> Stylistically the hotspot makefiles test a variable/expression
>>>>>>> against a
>>>>>>> value/constant not the other way around. So:
>>>>>>>
>>>>>>>  ifeq (undefined,$(origin OPENJDK_TARGET_CPU_ENDIAN))
>>>>>>>
>>>>>>> becomes
>>>>>>>
>>>>>>>  ifeq ($(origin OPENJDK_TARGET_CPU_ENDIAN),undefined)
>>>>>>>
>>>>>>> and the same for all the other comparisons you have added.
>>>>>>>
>>>>>>>
>>>>>>> Sorry.
>>>>>>>
>>>>>>> David
>>>>>>>