Dependency on developer tools

Scott Haneda talklists at newgeo.com
Fri Oct 23 13:30:12 PDT 2009


On Oct 23, 2009, at 1:11 PM, Ryan Schmidt wrote:

> On Oct 23, 2009, at 15:00, Scott Haneda wrote:
>
>> You just saddened me as I run multiple PPC servers :)
>
> I understand, and I am on your side. I want to support older  
> systems. But we appear to be in the minority.

I want to go out and buy all new Intel Xserves, but looking at those  
prices, and I am looking at Dell, blades, or build your own.   
Hopefully I can go FreeBSD and at least feel a little at home.  I am  
not looking forward to hardware compatibility issues, that is why I  
like serving on the Mac, I know that whatever hardware I plug into it,  
so long as that raid card says it works with OS X, I know it is going  
to work.

As long as we keep on top of the MAMP stack for MacPorts, I suspect at  
least I can get by.  In the end, I feel that developers of the  
upstream source should still target PPC and make it work.  Apple makes  
their tool chain in a way that it is possible, and I believe even  
desirable to support PPC.

If I was said source developer, I would actually want to fix it, even  
if zero people ran on PPC, it seems like a bug to me.

>> Do you have stats?
>
> I have no stats. I am recalling the recent situation with glib2: as  
> of MacPorts 1.8.0, the "-arch" flag is passed along to every port.  
> But glib2's behavior was such that if any "-arch" flag appears --  
> even just one -- it assumes you're building universal. This is  
> clearly a ludicrous assumption, but until now the developers  
> apparently did not encounter a situation where it was not true. If  
> you pass only a single "-arch" flag, as MacPorts now does when not  
> building universal, this somehow had the effect of making glib2  
> assume it was on Intel, always, regardless of what arch you actually  
> passed. glib2 would install successfully, but any port that needed  
> glib2, like say gtk2, would fail to build, with a weird message. It  
> took over a month and hundreds of lines of patches and scripts to  
> fix glib2. Other ports may have similar as-yet-undiscovered issues.

I do not know too much about this, but I do know that glib2 is  
probably a very widely used C lib, is that not correct?  How in the  
heck was this not noticed ages ago?  glib2 probably is in the top 10%  
of installed dependencies I would imagine.

>> I would be willing to wager there is a disproportionate amount of  
>> users that are on PPC using MacPorts.
>
> I don't doubt that there are still Mac users with PowerPC Macs using  
> MacPorts. But I am contending that the number of software developers  
> and port maintainers with PowerPC Macs is decreasing, meaning  
> developers and maintainers are more likely to introduce PowerPC-only  
> bugs that go undetected.

Weird.  If I ever get around to being a real developer :) I would just  
assume I had to have access to at least a few machines that I could  
use as a build farm, so to speak.  A G4 on ebay is around 125.00 or so.

>> I will bring this up also on the off list thread about Web site  
>> changes, as with a little plugging in of Analytics, or running the  
>> logs, we can find out exactly what the breakdown is, based on  
>> browser and OS.
>
> Sure, web site analysis could be interesting.

More on this in the off list thread...

> I also once suggested that MacPorts should send stats about what got  
> installed up to our server so we can tell what ports are popular,  
> and also what kinds of systems people are running them on. But this  
> was seen as "phoning home" and an invasion of privacy. It could be  
> an opt-in feature, but I thought this would make it less useful.

Ok, I can understand that.   I think we can get the birds eye view,  
since the stats will tell us which installer was downloaded.  We at  
the very least, know that there are installers for Snow, Tiger, and  
Leopard.  We know that Snow is all Intel, so the margin of error  
happens in the pre Snow downloads.

Adding those stats in which the user agent of the browsers that hit  
the url, we should be able to tell it it is PPC or Intel.  Would at  
the least be interesting.  Again, I will get pretty deep into this on  
the other thread.  I need to get an overview of what is running on the  
MP servers, and how it is all set up.

Data like this is important, in order to know how to better position  
the Web site for users. Look forward to being able to help out in an  
area where I think I am well suited to help out.

Have a good weekend.
-- 
Scott * If you contact me off list replace talklists@ with scott@ *



More information about the macports-dev mailing list