variant or platform

[Ryan Schmidt]

On Jul 15, 2007, at 14:25, Juan Manuel Palacios wrote:

> On Jul 15, 2007, at 3:11 PM, Ryan Schmidt wrote:
>
>> On Jul 15, 2007, at 10:03, Yves de Champlain wrote:
>>
>>>>> Mmm i386 is for intel, sure.  But x86 targets a wider set of  
>>>>> platforms (intel and AMD).  So I always saw i386 as a subset of  
>>>>> x86.
>>>>>
>>>>> Am I far off ?
>>>>
>>>>
>>>> On Jul 11, 2007, at 00:26, Yves de Champlain wrote:
>>>>
>>>>> I just saw "platform intel"  Does that really work ?
>>>>
>>>>
>>>> I really have no idea about "x86" or "platform intel". Where is  
>>>> the code in base that autoselects platforms? We could look there  
>>>> to see what it supports.
>>>
>>> it fetches uname, so whatever comes out is good
>>
>> Ah. Well, uname -a shows i386 on my MacBook Pro with Mac OS X, but  
>> I don't know what it outputs on all other operating systems on all  
>> other computers. So what do we do now?
>
> 	Part of the code that sets the platform name is in base/src/ 
> port1.0/portmain.tcl, but I have to admit I don't remember the  
> original question of this post in order to provide a helpful  
> answer :-S What is it that we're trying to agree on? An Intel Mac  
> appropriate platform name? intel? i386? something else?

Well, it was your thread: you were normalizing platform names and  
discovered things like "x86" and "intel". We were both wondering if  
"intel" was a valid platform name anywhere, and I don't know how to  
find out if "uname" returns "intel" on any computer. However, I can't  
find any port containing "platform intel" now. Did you change it  
already?

And you posited that "x86" worked on some platforms, while I know  
that "i386" is correct for Intel Macs. In terms of what happens if  
you write a platform selector in MacPorts, I don't think i386 is a  
subset of x86 at all; if I were to draw a Venn diagram, I don't think  
the two circles would intersect. uname either returns i386 (like on  
Intel Macs) or it returns x86 (on some other Intel computers). I  
think my point was just that any occurrences of "platform x86" or  
"variant x86" should *not* be changed to "platform i386" because the  
x86 selectors seem to have been used in the past to target non-Macs,  
which would make this software start to fail on Intel Macs.