[Rd] As a package author, is there a way to specify that your package is architecture (x86_64) specific?

[Simon Urbanek]

On Sep 9, 2011, at 10:30 AM, Steve Lianoglou wrote:

> Hi Simon, Prof. Ripley, and Dirk,
> 
> First: thanks again for the tips, it's great to have some of the "top
> bRass" providing this type of help.
> 
> Last (few) comments in line:
> 
> On Fri, Sep 9, 2011 at 9:41 AM, Simon Urbanek
> <simon.urbanek at r-project.org> wrote:
>> On Sep 9, 2011, at 8:36 AM, Prof Brian Ripley wrote:
>>> On Thu, 8 Sep 2011, Steve Lianoglou wrote:
> 
> [snip]
> 
> About the architecture thing:
> 
>>>> Ok, sorry for being imprecise. Let's see if we can figure out what it
>>>> is (more precise details are at the bottom of the email). I see x86_64
>>>> on every 64bit machine I touch (linux or mac), so I thought they were
>>>> interchangeable (as names).
>>> 
>>> Not so.  That's not what Windows uses (it mainly uses x64, sometimes amd64 or x86-64, almost never x86_64), and although it is what Solaris uses, your Linux x86_64 assembler (presumably GNU) will not work there.
>>> 
>> 
>> Moreover not all 64-bit machines are x86_64/amd64 - there is ppc64, Sparc, MIPS64, IA64, ... which was my point about this having to do with a very particular machine type (and my suspicion involving asm was correct ;)) and not 64-bit architectures in general.
> 
> Right, of course ... as soon as you mentioned ppc64, the distinction
> you folks were trying to point out was immediately clear, thanks.
> 
> [snip]
> 
> About the configure/autoconf testing thing:
> 
>>> You try to compile the crucial fragament of code in configure: there are lots of examples of that in the R sources (mainly in the m4 directory).
>> 
>> Yes, that is the right thing to do. It's only a few lines of configure.in if you want to use autoconf, but if you don't, it's still fairly easy in pure shell code, just a bit more legwork.
> 
> OK, thanks for that, too.
> 
> So, last question here -- say I catch this error condition during the
> configure (or shell) check code, and I realize that some bit of code
> won't work on this machine type.
> 
> How do I signal to R's build/compile process to "error out" on the
> package build proces, but just move on to the next arch/machine-type
> it should try to compile?
> 
> I mean: I can imagine how one could catch an error during the compile
> test and then define some var that an `#ifdef SOMETHING` could be used
> in my codebase to do one thing or the other, but in this case I'm not
> trying to direct the compiler down a different codepath that will work
> for the machine type (for now), I just want it to give up on the
> current build and try the build for the next machine-type in line.
> 

You can't. Your packages either builds or not. And as you should know by now there is no "next matching-type" in line since configure guarantees just one arch.

So you have two options:

a) fail on unsupported architecture. That means only if the architecture is supported there will be a binary. Simple and clean.

b) let configure detect the support and flag it. Then you can #ifdef your code and create a stub that simply calls Rf_error("unsupported architecture"). In fact in your particular case you could even possibly get away with #if __x86_64__ and no configure. It's less clean because the user doesn't know that your package doesn't work and may be clumsy if you have several R entry points.

Cheers,
Simon