[MacPorts] #44918: defect: boost @1.56.0_1+no_single+no_static+python27+gcc48 won't build on PPC Tiger (Mac OS X 10.4.11)

[MacPorts]

#44918: defect: boost @1.56.0_1+no_single+no_static+python27+gcc48 won't build on
PPC Tiger (Mac OS X 10.4.11)
------------------------------+--------------------------------
  Reporter:  Peter_Dyballa@…  |      Owner:  macports-tickets@…
      Type:  defect           |     Status:  new
  Priority:  Normal           |  Milestone:
 Component:  ports            |    Version:  2.3.1
Resolution:                   |   Keywords:
      Port:  boost            |
------------------------------+--------------------------------

Comment (by ryandesign@…):

 Replying to [comment:3 Peter_Dyballa@…]:
 > Replying to [comment:2 ryandesign@…]:
 > > The transcript above doesn't show a build failure, but somehow an
 extract failure of the llvm-3.3 port.
 >
 > No, no! You see me aborting after I saw that port is again going to
 build llvm-3.3. So that's quite OK.

 Oh, ok, I didn't realize that.

 > > I'm not sure if llvm-3.3 is expected to be able to build on PowerPC at
 this point. I'm not sure why llvm-3.3 is being used here either, but I
 suspect it's because you have the cctools port and/or the ld64 port
 installed with the +llvm33 variant?
 >
 > No, that's obviously not the case:
 >
 > {{{
 > port installed | egrep 'cctool|ld'
 >   cctools @806_3+llvm29 (active)
 >   cctools-headers @855_0 (active)
 >   dyld-headers @239.3_0 (active)
 >   ld64 @97.17_3 (active)
 >   openldap @2.4.39_0 (active)
 > }}}
 >
 > But ld64 looks suspicious, doesn't it?

 It doesn't look suspicious to me. I've checked on my PowerPC Tiger machine
 now, and it has the same installed.

 > > If so, you could avoid the sometimes problematic llvm dependency and
 its lengthy compilation time by reinstalling the cctools and ld64 ports
 without any llvm variant.
 >
 > To me it looks as if I have to choose one variant, so I'll try to build
 ld64 +llvm29 (and after checking whether any port depends on some
 particular LLVM version I'll try to upgrade cctools and ld64 to that
 version. At least I could build:
 >
 > {{{
 >   llvm-2.9 @2.9_14 (active)
 >   llvm-3.0 @3.0_13 (active)
 >   llvm-3.1 @3.1_8 (active)
 >   llvm-3.2 @3.2_3 (active)
 > }}}

 On Intel machines running OS X 10.5 or later, you must select one of the
 llvm variants. On PowerPC machines running any OS X version or on Intel
 machines running OS X 10.4 it is permitted to not select any of them. But
 that was already the case for you.


 Replying to [comment:5 cal@…]:
 > The debug output shows that the gcc48 variant did run, and it also shows
 that MacPorts chose the clang-3.3 compiler for boost. Installing llvm-3.3
 is the obvious consequence, trying to install a different llvm won't help
 here – the problem is the compiler choice of the boost port. You were
 right in assuming the boost Portfile is to blame.

 Right, I just realized that as well. And llvm-3.3 doesn't build on
 PowerPC; I found the ticket: #39849. It was closed as not to be fixed. So
 it seems there is no compiler in existence that can build the current
 version of boost on a PowerPC Mac.


 Replying to [comment:7 Peter_Dyballa@…]:
 > What is then the purpose of packages like gcc_select, llvm_select,
 clang_select?

 To assist you in more easily compiling your own software outside of
 MacPorts. MacPorts itself is not supposed to take any notice of what you
 select (although some ports still erroneously do, and need to be fixed not
 to do that).

-- 
Ticket URL: <https://trac.macports.org/ticket/44918#comment:8>
MacPorts <http://www.macports.org/>
Ports system for OS X