building NEURON on OSX 10.4 against X11

avicenna · Post by **avicenna** » Thu Sep 06, 2007 4:18 pm

Same results with 6.1.931; nrngui displays a menu over ssh on my Linux machine but not on the Mac machine itself. idemo does work on the Mac machine, however.

Post by **hines** » Thu Sep 06, 2007 5:14 pm

Although I configured, built, and installed iv as you previously described, I can't run idemo because its linking assumes that libUnidrawhines.3.dylib and libIVhines.3.dylib are located in /usr/lib. nrniv, however, correctly assumes that said libraries are where I installed them (i.e., in /tmp/neuron/i686/lib).

That is strange. On my machine
otool -l idemo
gives
...
Load command 7
cmd LC_LOAD_DYLIB
cmdsize 80
name /Users/hines/neuron/ivx11/i386/lib/libIVhines.3.dylib (offset 24)
time stamp 1188860978 Mon Sep 3 19:09:38 2007
current version 4.3.0
compatibility version 4.0.0

which is consistent with the prefix I used

Post by **hines** » Thu Sep 06, 2007 5:22 pm

My machine is (from "about this mac") 10.4.10
2 GHz intel core duo

in iv/src config.guess yields
i386-apple-darwin8.10.1
which is, itself, a puzzle but is apparently what apple, autoconf
presently intend. And gcc --version gives
i686-apple-darwin8-gcc-4.0.1 (GCC) 4.0.1 (Apple Computer, Inc. build 5250)

avicenna · Post by **avicenna** » Fri Sep 07, 2007 11:17 am

My machine is has a 2.16 GHz Core 2 Duo and is running 10.4.10; iv's configure identifies the system type as i686-apple-darwin8.10.1.

Post by **hines** » Fri Sep 07, 2007 2:07 pm

I've uploaded a new
http://www.neuron.yale.edu/ftp/neuron/v ... -17.tar.gz
both iv and nrn should have a config.guess of i386 on the newer macs (why, I don't know).

avicenna · Post by **avicenna** » Mon Sep 10, 2007 5:51 pm

The iv config.guess script still reports my system as being i686-apple-darwin8.10.1. I tried rebuilding it along with NEURON 6.1.933; the same results as previously noted were observed.

Post by **hines** » Mon Sep 10, 2007 7:04 pm

try downloading again. Every one of my versions of autoconf except the one on the mac generate a config.guess that returns i686 instead of i386. Either one is ok but it should be the same for both iv and nrn. i386 is useful since the configure for mpich2 on the mac also returns i386
in case you want to make use of the dual cores.

avicenna · Post by **avicenna** » Tue Sep 11, 2007 4:10 pm

Now iv's config.guess returns i386-apple-darwin8.10.1 while the config.guess in NEURON 6.1.933 returns i686-apple-darwin8.10.1 (I erased and downloaded both tarballs).

Post by **hines** » Tue Sep 11, 2007 5:24 pm

Well then, I guess that was a wild goose chase. I re-uploaded an iv-17.tar.gz that
has a config.guess that returns i686.I was thinking that on your machine there were incompatible config.guess for iv and nrn. As long as they are both the same, that should not be a problem.

I must say that I am at a loss as to what is the cause of the missing nrniv windows under x11 on your machine. One way
to transcend the difficulty is to build iv and nrn with the --enable-carbon configure options. The only way I know to diagnose the x11 problem is to start adding printf statements in the code path between the hoc statement
objref g
g = new Graph()
and when the x11 window is mapped to the screen in the iv/src/lib/IV-X11/window.cpp file.

avicenna · Post by **avicenna** » Mon Sep 17, 2007 12:41 pm

Compiling the latest alpha release of NEURON against Carbon seems to work fine.

I tried building NEURON against X11 on another Intel Mac running MacOSX 10.4.10 and Apple's X11 server with the same results I obtained on my iMac.

I'll see if I can discover anything with your debugging advice.