[tex-live] small bug re installing x86_64-darwin vs. universal-darwin

[Adam R. Maxwell]

On Jan 13, 2013, at 14:37 , Mojca Miklavec <mojca.miklavec.lists at gmail.com> wrote:

> On Sun, Jan 13, 2013 at 4:44 PM, Adam R. Maxwell wrote:
>> On Jan 13, 2013, at 00:29 , Mojca Miklavec wrote:
>> 
>>> - 10.6 could be booted in either 32-bit or 64-bit kernel; uname was
>>> thus returning different results depending on how one booted the Mac.
>>> If one relied on uname only, without testing for
>>> "hw.cpu64bit_capable", TeX Live would suddenly stop working when
>>> booting into a different kernel. But the code above is immune to that.
>> 
>> FWIW, I think the kernel is irrelevant here, as the 32-bit kernel can
>> still run x86_64 code (and vice versa).
> 
> [I don't want to go too off-topic, but ...]

Me neither, but I have TL questions...

> Not entirely. Yes, it can run 64-bit code, but uname returns a
> different value depending on which kernel is loaded. Consequently the
> output of config.guess script is almost useless (and would have to be
> overloaded/fixed later in perl code), in particular when it returns
> i386-apple-darwin10.* (when it says that one is using 32-bit kernel on
> 10.6).

My point, probably poorly stated, was that if you can reboot into a
different kernel, as your statement requires, then you must have 64-bit
hardware.  Consequently, I don't see any way that TeX Live can "stop
working" due to such a reboot.  I'm probably missing something…

Does TL use config.guess during install?  That seems wrong on the Mac
unless you're actually compiling, since it makes its guess based on a
C compiler which likely isn't even available.  Your heuristic looks
more reasonable for average users.

Adam