Review OSX universal mode patch

[Mike Swingler]

The patch looks fine to me. The value of "universal" is required to let the build infrastructure pass both architectures to the compiler, and let the compiler do the work of running twice, and then combining the results.

The #ifdef is done so that in the separate runs of the compiler, one value is hardcoded, and the other is not. At runtime, the expected values are output, and there is a single Mach-o executable on disk, with two architecture slices inside of it.

Perhaps David was unfamiliar with the Universal build process, but this is the exact sort of impedance mis-match that comes from the mainline developers not eating, breathing, and living on the configuration in question. The individual who signs up to support 32/64 Universal, is essentially signing up to defend changes like this, and whomp the mainline developers over the head when they break the unsupported configuration. Not an easy task.

Regards,
Mike Swingler
Apple Inc.

On May 10, 2012, at 10:17 AM, Julien Ponge wrote:

> Ok, so on pure technical grounds how about the patch that Henri initially offered for review?
> 
> We are talking here about the "first" step that Charles mentionned. 
> 
> Cheers
> On jeudi 10 mai 2012 at 18:26, Mike Swingler wrote:
> 
>> On May 10, 2012, at 9:20 AM, Julien Ponge wrote:
>> 
>>>> Frankly, it seems a little silly to me to expect that anyone would
>>>> even be capable of maintaining 32/universal OpenJDK builds for OS X
>>>> unless you make the first step of making it *possible* to build
>>>> 32/universal OpenJDK for OS X. Making them buildable seems like a
>>>> first step, eh?
>>> 
>>> +1
>>>> * Is there community interest in having 32/universal build support in
>>>> OpenJDK on OS X? I think the answer is obviously yes.
>>>> * Has anyone stepped up to provide patches to make 32/universal builds
>>>> possible on OS X? The answer is also yes.
>>> 
>>> Henri has done a fantastic job by making it possible to use Java 7 on the Mac, way before we could get any "official" pre-release of any sort.
>>>> * Does OpenJDK work in 32-bit or universal modes on OS X? "Works" is
>>>> hard to answer definitively, without running the TCK, but *none* of
>>>> Henri's builds (which you praised openly at FOSDEM, Dalibor) have run
>>>> TCK either, right? So it "works" as well as any build of OpenJDK that
>>>> hasn't run TCK, which is very likely *all* community builds in the
>>>> wild.
>>> 
>>> Yes, I think most of us do not care about the build passing the TCK or not. We use such builds for development, hence "fresh" is better than "blessed".
>>>> openjdk-osx-build is a valuable resource for any of us working on OS
>>>> X, and it would be a terrible shame to lose it.
>>> 
>>> Unfortunately it seems like it has already been lost...
>> 
>> Let's be clear here. Nothing is lost unless it's unmaintained.
>> 
>> If someone with the necessary skills is willing to commit here and now to supporting 32-bit into the future, I'd support that. But without that commitment, there is no point to re-introducing 32-bit support, only to let it rot.
>> 
>> So, any takers?
>> 
>> Mike Swingler
>> Apple Inc.
>