[ADMB Users] MacOS progress/wishlist

[Johnoel Ancheta]

For 2., It's probably easier not to use -m32/64.  If someone needed
32 they could just

$ ./configure CXXFLAGS=-m32

Is this okay?

For 4., I'll fix the test target to avoid some builds.

On 6/15/11 11:50 AM, Ben Bolker wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 11-06-15 08:51 PM, Johnoel Ancheta wrote:
>> On 6/15/11 5:56 AM, Ben Bolker wrote:
>>>     Thanks to Dave for the response yesterday.  I have done some
>>> struggling on my own, with the following comments/wishlist items:
>>>
>>> 1. it's still really easy to find the old ADMB SVN repository and get
>>> confused, at least it was for me.
>>>
>>>     Could someone please update
>>>
>>> http://code.google.com/p/admb-project/source/checkout
>>>
>>>     so that it indicates that it is an old version and points to (or
>>> replicates the information on)
>>>
>>> http://admb-project.org/news/admb-anonymous-source-version-control-available
>>>
>>>
>>>     ?
>> Can't change it...
>>> 2. Examining GNUmakefile showed "-m32" in the compiler flags.  Changing
>>> it to "-m64" followed by "make clean; make; make install" gets me
>>> partway through my problem.
>>> I'm not sure what the best solution is in terms of helping future users
>>> get through this.  I poked around a bit in
>>> scripts/configure/configure.ac , which is ultimately what determines the
>>> contents of GNUmakefile; there is apparently (?) logic that is trying to
>>> autodetect 32- vs 64-bit systems, but I guess it fails on my system?
>> $ ./configure BUILD_64BIT=yes
>    Thanks.  Any ideas about how to have this detected
> correctly/automatically?
>
>>> 3. after that I ran into this problem:
>>>
>>> ld: duplicate symbol lgam(double)in
>>> /usr/local/admb/lib/libado.a(cgamdev.obj) and
>>> /usr/local/admb/lib/libdf1b2o.a(df13tester.obj)
>>>
>>>
>>> which is referenced here:
>>> http://groups.google.com/group/admb-users/browse_thread/thread/c90d220c6b6b17ca
>>>
>>>
>>>
>>>     My solution was:
>>>
>>>     a. delete src/df1b2-separable/df13tester.cpp
>>>     b. remove the df13tester.obj line from src/df1b2-separable/objects.lst
>>>     c. remove object files containing the bad stuff from the build
>>>
>>> find . -type f -exec grep -H df13tester {} \; | grep -v svn
>>>
>>> although "make clean" would probably have been easier
>>>
>>>     d. make; make install
>>>
>>>    this allows me to compile the file I am interested in.
>>>
>>>     *** It would be really nice if one of the developers could make these
>>> changes (tested appropriately of course) to the SVN source ... ***
>> I'll look into it.
>   Thanks.
>
>>> 4. "make test" still fails on my machine, starting from here:
>>>
>>> g++ -c -I/usr/local/gtest-1.5.0/include -O3 -Wno-deprecated -D__GNUDOS__
>>> -Dlinux -DSAFE_ALL -DUSE_LAPLACE -fpermissive -I.
>>> -I/usr/local/admb-svn2/build/admb-10.1-macos10.6.7-xcode3.2.6-32bit/admb-10.1-macos10.6.7-xcode3.2.6-32bit/include
>>>
>>> test_atan2.cpp
>>> test_atan2.cpp:1:25: error: gtest/gtest.h: No such file or directory
>>> test_atan2.cpp:13: error: ‘::testing’ has not been declared
>>>
>>>     any ideas?
>>>     (this is lower priority for me now since I can compile my TPL file of
>>> interest)
>> Use
>> $ make verify instead
>>
>    ... if "make test" doesn't work/isn't expected to work perhaps it
> would be a good idea to remove the target ... ?
>
>>> 5. It would be really nice, and I think it is possible, to set up ADMB
>>> so that it could generate "multi-architecture" TPL files.  For example,
>>> when I query the results of my compilation I get
>>>
>>> glmmadmb: Mach-O 64-bit executable x86_64
>>>
>>> I think it is possible to use something like CXXFLAGS="-arch i386 -arch
>>> x86_64" at some point, but when I naively hack GNUmakefile to reflect
>>> this I get
>>>
>>> lipo: /var/folders/5A/5A99Opv1HbSJQA2Bnw6oqE++r-g/-Tmp-//ccERu1s8.out
>>> and /var/folders/5A/5A99Opv1HbSJQA2Bnw6oqE++r-g/-Tmp-//ccOGn9us.out have
>>> the same architectures (x86_64) and can't be in the same fat output file
>> ???
>>>     I am most definitely not a Mac expert.  Does anyone know how to do
>>> this (by hacking GNUmakefile if possible, even better would be
>>> specifying arguments to ./configure) ?
>>>
>    thanks ...
>       Ben
>
>
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.10 (GNU/Linux)
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
>
> iEYEARECAAYFAk35KQoACgkQc5UpGjwzenOibQCghqAKc/15v+Wc7GdJQsEfXLiG
> nwMAn3Qv0KR6J4YYJNRmRopcnApLLovg
> =iacJ
> -----END PGP SIGNATURE-----