Re: Patch for config.guess for "cross-compiling" x86_64 -> powerpc on Mac OS X <= 10.6

[Mojca Miklavec]

Dear Ben,

On 23 April 2017 at 10:59, Ben Elliston wrote:
> On Sat, Apr 22, 2017 at 11:08:37PM +0200, Mojca Miklavec wrote:
>
>> (1) Which platform am I building for?
>>  (2) Which platform am I running on / which binaries should I download?
>
> The answer is (2) -- which platform am I running on.
>
>> I've got quite some criticism and some approvals, but no answer from
>> the maintainer yet. So: is there any chance to accept the patch?
>
> Furthermore, I'm not happy to accept patches that further entrench the
> use of CC_FOR_BUILD, sorry.

In this particular case the compiler has already been consulted once with

    if (echo '#ifdef __LP64__'; echo IS_64BIT_ARCH; echo '#endif') | \
        (CCOPTS="" $CC_FOR_BUILD -E - 2>/dev/null) | \
        grep IS_64BIT_ARCH >/dev/null


to determine that this is not x86_64. Is it a problem to ask it once
more? (And why?)

I understand that calling the compiler will be a problem on, say, OS X
10.9, where this asks the user to install additional gigabytes of
software (Xcode), but in this particular case there's really no harm
being done by an additional call.

Do you have any example of a well-written configuration of some open
source software which needs to know all the info (for example in order
to include some code written in assembler for that particular
architecture) and does it without consulting config.guess? Something
that should work out of the box when I use CC="gcc -arch ppc"?

I have problems compiling at least three pieces of different open
sources software because of this issue.

But in case you *really* want to avoid ever using CC_FOR_BUILD: why is
it even called for Darwin?

Mojca