Thanks much for the pointer to distfiles; it does appear to have what I was looking for. I have a fairly new installation here so playing around & changing Macports installations is still an option.

I have heard about possible conflicts with Homebrew. I might toss it as I don't have much installed with it either.

If I find out something of interest I'll report back. but Xmas is in the way so it may take a few days.

Uli


On 12/24/16 11:13 AM, Ryan Schmidt wrote:
Please keep threads on the mailing list by replying to all.


On Dec 23, 2016, at 12:59, Uli Wienands wrote:
On 12/23/16 10:53 AM, Ryan Schmidt wrote:

On Dec 23, 2016, at 10:36, H.-Ulrich (Uli) Wienands wrote:

I am trying to install octave on a Mac Pro running 10.8.(latest). I have Xcode 
5.1.1 installed.

A bunch of deps is installed successfully but then it chokes.

I finally track this down to an inability to install any compilers (? that is 
what it appears to be).

Trying to build up from the bottom I install llvm 3.4, 3.5 etc. Works ok until 
I hit llvm-3.9. It appears llvm-3.
9 has itsels in its list of dependencies.

I made sure my Macports installaion was up-to-date (I installed it only a 
couple of weeks ago but still I updated everything) and tried to clean whatever 
would not install. No joy.

Here is the tail from the log of my atttempt to install llvm-3.9. Note that none of the 
clangs would install. And I never get an error message beyond "Installatino of 
..." has failed. I also tried installing gcc46 (which I have installed even on a 
fairly ancient on a 10.5 PPC system, no joy either.

Any hint would be appreciated.
We do have a ticket about that:

https://trac.macports.org/ticket/53138
Well, in my experience this problem is much more widespread than clang-3.x. As 
I said below, I cannot even install gcc46 (which should not require any of the 
clangs, and in fact on my Leopard system did not require these).
#53138 is about circular dependency problems with some clang ports on some 
systems. If you are having trouble installing gcc46 that might be a different 
problem. Or it could be the same problem, since gcc46 does depend on other 
ports, one or more of which might blacklist certain compilers on certain OS 
versions which might result in an attempt to install a MacPorts clang on 
certain OS versions.


I looked for Macports 2.3.4 (which works according to the note in the ticket) 
but I can only find the sources, no more installers :-(.
It is not useful for you to attempt to install MacPorts 2.3.4 at this point, 
but all of our installers are at:

https://distfiles.macports.org/MacPorts/


I am not at all sure how to install this without an actual installer (and, come to 
think of it, I only have a Homebrew gcc6 that works; no idea this would be enough 
to compile & install Macports form sources).
There is no need for you to attempt to build MacPorts from source, but should 
you wish to do so, the compilers provided by Xcode and its command line tools 
are the ones you should use. You should not attempt to use FSF GCC.

You should not have Homebrew (or anything installed with it) installed while 
using MacPorts; they can conflict with one another. Use only a single package 
management system and install the other(s).


So I feel pretty much stranded, as I would assume others do as well. How about 
reverting to a working version of Macports until this can be sorted out? This 
seems to me a fairly serious issue.
MacPorts 2.3.5 itself is probably not the problem. There were not many changes 
between 2.3.4 and 2.3.5. It is more likely that there is a problem with a 
particular port or ports that causes a circular dependency to be experienced. 
The ports involved with the circular dependency chain need to be examined to 
determine where the chain can be broken.

It is easy for port developers to overlook this kind of problem when committing 
changes to ports, because the developers already have the dependencies 
installed on their systems. It is only new users who don't have the 
dependencies installed yet who encounter the problem when MacPorts is unable to 
resolve the circular dependency chain.


Reply via email to