[MacPorts] #48178: circular dependency apple-gcc42/xpm, and many others
MacPorts
noreply at macports.org
Fri Jul 10 08:30:42 PDT 2015
#48178: circular dependency apple-gcc42/xpm, and many others
--------------------------+--------------------------------
Reporter: herzog@… | Owner: macports-tickets@…
Type: defect | Status: new
Priority: Normal | Milestone:
Component: ports | Version: 2.3.3
Resolution: | Keywords:
Port: apple-gcc42 |
--------------------------+--------------------------------
Comment (by herzog@…):
A change to Xcode 2.5 would be a problem for me, since several of my
projects building
very well with Xcode 2.4.1 don't with Xcode 2.5. Mainly because of
a change in
Interface Builder.
Moreover, according to Apple's release notes, Xcode 2.5 cannot be
used to develop
for Mac OS 10.5. A change to Xcode 3 seems to be impossible, too,
since it doesn't run
with Mac OS 10.4.
Nevertheless, I have tried to use Xcode 2.5 running the system
from an external drive,
and the problem remains.
So I returned to Xcode 2.4.1 and tried to build apple-gcc42
manually following the hints
in the file
{{{
$(port work apple-gcc42)/gcc-5666.3/README.Apple
}}}
I created a folder
{{{
testplace,
}}}
copied the archive 'gcc-5666.3.tar.gz' into a subdirectory of my
home directory, expanded
it into a folder named
{{{
gcc-5666.3,
}}}
changed into this folder and called
{{{
mkdir darwin
cd darwin
../configure --prefix=/testplace
make bootstrap
make install
}}}
Already the configure script is a problem. It assumes that the
minimum Mac OX is system 10.5 and
that the source root is
{{{
/Developer/SDKs/MacOSX10.5.sdk
}}}
Such a source root directory doesn't exist, neither with Xcode
2.4.1 nor with Xcode 2.5. So one
shouldn't be worried that compilation fails. I tried to call
configure with the option
{{{
--with-build-sysroot=/Developer/SDKs/MacOSX10.4u.sdk
}}}
Actually, I added further options to be as close as possible to
the call done by the apple-gcc42 port:
{{{
--prefix=/testplace
--enable-languages=c,c++,objc,obj-c++
--with-system-zlib
--disable-nls
--with-slibdir=/usr/lib
--with-gxx-include-dir=/usr/include/c++/4.0.0
--host=powerpc-apple-darwin8
--target=powerpc-apple-darwin8
}}}
In addition, I had to create a directory
{{{
testplace/powerpc-apple-darwin8/lib
}}}
to avoid 'no such directory' errors from the bootstrap script and
a hard link
{{{
testplace/powerpc-apple-darwin8/include
}}}
pointing to /usr/include, for otherwise, after a time of
compilation, I got complains that
c standard headers like 'errno.h' and 'time.h' were missing. I am
not sure whether the link
points to the correct location. Maybe it should point to a
location inside the '/opt'
directory containing these headers?
Finally, at the end of the build process for
'libgcc_s.1.dylib.tmp', the linker reports
{{{
/usr/bin/ld: -syslibroot: multiply specified
}}}
and stops execution of the bootstrap script. Indeed one finds as
arguments to the linker both
{{{
-isysroot /Developer/SDKs/MacOSX10.4u.sdk (specified
twice)
and
-isysroot/Developer/SDKs/MacOSX10.5.sdk
}}}
This seems to be a build or configure script or Makefile bug.
Since I have no idea about the gcc
build mechanism I even don't know where to look for it.
All the actions above are assembled in the script
{{{
build_gcc1.xc
}}}
and its output in the text files
{{{
configure1Log.txt
bootstrap1Log.txt
}}}
(attached to this report).
--
Ticket URL: <https://trac.macports.org/ticket/48178#comment:8>
MacPorts <https://www.macports.org/>
Ports system for OS X
More information about the macports-tickets
mailing list