Hello,
There was a bug similar to what you observed once upon a time, but it
should have been fixed a release or two (or three) ago. I just tried a
fresh install of 1.0.15 on a Linux box (but Fedora, but that should not
matter) and used the CCPN provided Tcl/Tk and Python, and it gave the
correct paths in the analysis script:
#!/bin/csh
setenv CCPNMR_TOP_DIR /home/wb104/ccpnmr
setenv PYTHONPATH .:$CCPNMR_TOP_DIR/ccpnmr1.0/python
setenv LD_LIBRARY_PATH /home/wb104/ccpnmr/tcl8.4/lib:/home/wb104/ccpnmr/tk8.4/lib
setenv TCL_LIBRARY /home/wb104/ccpnmr/tcl8.4/lib/tcl8.4
setenv TK_LIBRARY /home/wb104/ccpnmr/tk8.4/lib/tk8.4
/home/wb104/ccpnmr/python2.4/bin/python -i -O $CCPNMR_TOP_DIR/ccpnmr1.0/python/ccpnmr/analysis/AnalysisGui.py $1 $2 $3 $4 $5
so I'm a bit puzzled what is going wrong in your case. (I haven't yet
tried the interrupted problem.)
On the question of whether to have an installation on one computer or on
each users, that is a tough question. In Cambridge we have done both (but
the "one" computer is actually two installations, one for 32-bit and one
for 64-bit Linux boxes).
The main problem with the cental installation is that it generally means
that one person is responsible for updating the code, and if they are away
it can be a pain for everyone else. (As it happens, this happened in our
lab today because I was away waiting for my broadband box to arrive at
home, which fortunately it eventually did.) There are of course ways
around this (e.g. have more than one person know the appropriate password
-- which eventually is what resolved the issue here today).
There are a couple of problems with the other approach. One is that users
have to be comfortable updating their own version (and hopefully most
users are). Another issue (and one that is not totally resolved) is if
two users want to pass a project back and forth (of course you cannot work
concurrently on Analysis, I mean serially). If Alice has release X and
Bob has release Y > X, then with our current methodology, anything Alice
saves can be read by Bob, but it's possible that something Bob saves might
not be readable by Alice (it depends on whether the data model changed
between X and Y, which quite commonly it does). With the new API we are
hoping to be able to address this issue. The workaround here is to just
get Alice to upgrade to release Y, so it's not the end of the world.
Note that we are also about to undergo a major release upgrade, so that
might make the central installation option more appealing right now. (We
hope the upgrade will be fairly easy, but we haven't worked out all the
details yet.)
Wayne
On Fri, 17 Aug 2007, Patrick van der Wel wrote:
> Hello,
>
> I downloaded the ccpnmr/analysis package yesterday (v.1.0.15) to install
> it on a different system (running Suse Linux 9.2, I think). During the
> installation (where I installed it in the user account rather) I had to
> use the separate downloads of Tcl, Tk, etc.
>
> Both during the installation itself and the generated program scripts I
> noticed an apparent bug regarding the TK locations. Inside the analysis,
> updateCheck etc scripts I find the following library locations:
>
> export TCL_LIBRARY=/home/usr/software/ccpnmr/tcl8.4/lib/tcl8.4
> export TK_LIBRARY=/home/usr/software/ccpnmr/tcl8.4/lib/tk8.4
>
> Note that the second one is wrong.
>
> Something similar occurs during the execution of the installCode python
> script. If I uncompress and compile the local TK version, interrupt
> installation afterwards and restart, it does not find the unpacked TK
> code and I needed to do it all over again (without interruption it works
> ok, except for the final scripts listed above)
>
> Patrick
>
> PS. Another question: the way the updates are done (requiring
> permissions) it does not seem advisable to install it systemwide for a
> range of users on the same computer. Do people typically install a
> version for each user separately, or is there a better way around it?
>
|