O.k. Wayne,
Thanks. I just talked to Brian and he mentioned the install script bug.
Also we (and I said we were idiots) fundamentally missed the point that
ccpnmr analysis _includes_ the api and formatconverter. No wonder we
got a little confused. Mea culpa. We will give this another go, and try
another machine as well. Probably next week now as I have less
interesting (less fun) things to do, like interfacing with Microsoft
Word. Don't worry, we'll get there.
Keep up the good work.
Cheers,
Paul
On Fri, 27 Feb 2004, Wayne Boucher wrote:
> This whole installation thing can be a bit of a nightmare. When we first
> shipped this code we had binaries for everything, assuming Linux was bound
> to be good enough to handle that, but for many of the Linux computers we
> tried the libraries/executables just did not work (in particular, that
> first distribution did not even work on my own computer because it had
> been compiled on Tim's and his C libraries were newer than mine). It is
> shared libraries that seem to cause the most problems, they can compile in
> ways that look perfectly harmless (and work) but at runtime they can
> easily fall over on another computer. (The reason we used Tim's computer
> on that first distribution was because my Python libraries had a package
> called Tix compiled in, and unless the Tix libraries had also been shipped
> those libraries would not work on anybody's computer who did not have
> Tix.) So this is why we went to compilation, which is always bound to be
> hard, and in particular is not going to just be a simple "make" because of
> all the third-party software required. If we had enough time we could try
> to put something like "configure" together so that we could try and figure
> out what was where on the host computer, but my guess is that in most
> cases the user would have to at the minimum compile Python (with Tk) and
> Mesa (or download header files and libraries for OpenGL).
>
> We have three releases (API, FormatConverter and Analysis) because some
> people might only want the API (but I suspect that is nobody so far),
> some people might only want the FormatConverter (and that is most people
> so far) and some people might also want Analysis. The API release is a
> subset of the FormatConverter release which is a subset of the full
> release, and if you want the full release then do not download either of
> the other two releases on top of that because it will mangle a few crucial
> files (to do with the API). The bottom line is that you should only be
> downloading one of the releases.
>
> Anyway, onto the actual problems you experienced. The easiest problem to
> fix is the "comp_arch" one, that is a bug in the installation script which
> Brian found a few days ago and is now fixed in a release which went out
> late yesterday afternoon (I'll separately email you that). That would
> have certainly scuppered the rest of that script (unless your computer guy
> knew enough about Python to figure out how to fix it). (It's a relic of
> the change from Mesa 4 to Mesa 6, since the Mesa team decided to make
> compilation a lot harder than it used to be.)
>
> The other problems are hard to fathom. I don't think I've ever seen Tk
> fail to compile so something awful is happening on your computer. (The
> error messages reported for tk3d.o do no seem to correspond to lines in
> the Tk source code that I have where those error messages would be
> plausible.) And complaints about tkDecls.h are truly wierd, that is so
> fundamental to Tk that errors there are definitely the end of the world.
> (That tkDecls.h is not the one included in the release but it ought to
> work.) It's almost as if the compiler is picking up the wrong platform.
>
> As well as the install script there is an installation note called INSTALL
> in the top-level directory, which might help, for example, that mentions:
>
> "To get compilation of Python to work with the distributed versions
> of Tcl/Tk you might need to set the following environment variables
> first:
>
> setenv LD_LIBRARY_PATH
> ${CCPNMR_TOP_DIR}/tcl8.3/lib:${CCPNMR_TOP_DIR}/tk8.3/lib
>
> where CCPNMR_TOP_DIR is is the ccpnmr directory created by the
> unpacking of the tar file."
>
> (It is hard to know exactly which environment variables definitely need to
> be set because all our machines are polluted so badly that something could
> be sneaking in the back door without us knowing it.)
>
> As with all documentation, I'm sure the installation notes could be
> expanded and clarified and hopefully eventually it will get there. (It's
> useful having reports from people having problems for exactly this
> reason.)
>
> The Tkinter error ("No module named _tkinter") implies that the Tk bits of
> Python were not successfully compiled. (Which I guess is at least
> consistent with the fact that you were getting Tk compilation errors.) In
> fact after you have compiled Python (or if you intend to use your own
> version) the best way to check that it will work at all with Analysis (or
> the FormatConverter) is just to type in a shell:
>
> python
> >>> import Tkinter
>
> (use the full path for the executable if you have several on your
> machine) and if that fails (which it will for your installation as it has
> been described in your email) then something is wrong. The installation
> notes describe what to do to get Tk compiled in to the distributed Python
> (you have to edit some file, the installation script is supposed to do
> this all for you).
>
> Wayne
>
> On Fri, 27 Feb 2004, Paul Driscoll wrote:
>
> > Hi Guys,
> >
> > We have not gotten so far as Brian....
> >
> > Yesterday we tried to get going with ccpnmr really for the first time:
> > sorry it has taken so long but we have been short-handed.
> >
> > We regard ourselves as the sort of beta-testers who represent the average
> > (idiot?) users who want to try the software but are going to need a fair
> > bit of hand-holding. Our local systems manager, being much more up-to-speed
> > with these things than I, was on hand as we tried to install the packages
> > from the website - downloaded yesterday - both over the top of his
> > pre-installed python, mesa, tk, tcl, etc., and by explicit incorporation of
> > the packages supplied. Basically we could not get things working (details
> > below) but perhaps more importantly we would not have gotten as far as we
> > did were it not the some expert help. As a relatively naive computer
> > user, I was struggling to understand the order in which things should be
> > installed, when environment variables should be set, etc. Would it be
> > possible to reiterate how the installation should best be performed from
> > scratch (as a complete item, i.e. api, format and analysis). Though there
> > is some guidance in each of the individual Install scripts (which overwrite
> > themselves when unpacked, which I found confusing) it would be good to have
> > some overarching recipe for 'naive' compilers. Maybe this already exists
> > and I have simply missed it?
> >
> > Cheers,
> > Paul
> >
> > Details....
> >
> > [Summary from my systems manager trying to install in his desktop:]
> >
> > It appears that neither the supplied tk nor mesa will compile on my machine
> > (Redhat 9.0, which has kernel 2.4.24, python-2.2.2-26, python-devel-2.2.2-26,
> > tk-8.3.5-88, tcl-8.3.5-88, XFree86-Mesa-libGLU-4.3.0-2,
> > XFree86-Mesa-libGL-4.3.0-2). As a result analysis will not compile.
> > I cannot compile analysis using the preinstalled versions of python,
> > mesa, tcl and tk.
> >
> > Errors while trying to install the analysis bit (it was actually compiling tk):
> > {standard input}: Assembler messages:
> > {standard input}:663: Error: symbol `dy' is already defined
> > {standard input}:669: Error: symbol `dx' is already defined
> > {standard input}:716: Error: symbol `q' is already defined
> > {standard input}:722: Error: symbol `p' is already defined
> > {standard input}:728: Error: symbol `q' is already defined
> > {standard input}:734: Error: symbol `q' is already defined
> > make: *** [tk3d.o] Error 1
> >
> > and while compiling mesa:
> > Traceback (most recent call last):
> > File "installCode.py", line 514, in ?
> > main()
> > File "installCode.py", line 475, in main
> > compileMesa(comp_arch)
> > UnboundLocalError: local variable 'comp_arch' referenced before assignment
> >
> > while compiling analysis itself, lots of parse errors ending with:
> > /usr/include/tkDecls.h:1082: parse error before '}' token
> > /usr/include/tkDecls.h:1087: parse error before '*' token
> > make: *** [py_contour_file.o] Error 1
> >
> > and while trying to test analysis:
> > Run Analysis (as test)? y
> > Traceback (most recent call last):
> > File
> > "/d/mckenzie/linux/ccpnmr/ccpnmr/ccpnmr1.0/python/ccpnmr/analysis/AnalysisGui.py",
> > line 34, in ?
> > import Tkinter
> > File
> > "/d/mckenzie/linux/ccpnmr/ccpnmr/python2.2/lib/python2.2/lib-tk/Tkinter.py",
> > line 35, in ?
> > import _tkinter # If this fails your Python may not be configured for Tk
> > ImportError: No module named _tkinter
> > >>>
> >
> > ++++++++++++++++++++
> >
> > *** PLEASE NOTE THAT DUE TO UNSOLICITED MESSAGE OVERLOAD
> > *** MY ADDRESS IN FUTURE WILL (ONLY) BE:
> > *** [log in to unmask]
> > *** PLEASE CHANGE YOUR ADDRESS BOOK ACCORDINGLY.
> > *** THANK YOU.
> >
> > ++++++++++++++++++++
> >
> > Dr Paul C. Driscoll
> > Reader in Structural Biology
> > Joint UCL/LICR NMR Laboratory
> > Bloomsbury Centre for Structural Biology
> > Department of Biochemistry and Molecular Biology
> > University College London
> > Gower Street
> > London WC1E 6BT
> >
> > Office (answer)phone: (44)-20 7679 7035
> > Department fax: (44)-20 7679 7193
> > E-mail: [log in to unmask]
> > Web: http://www.biochem.ucl.ac.uk/~driscoll
> >
>
--
*** PLEASE NOTE: DUE TO PROBLEMS OF UNSOLICITED MESSAGE OVERLOAD
*** MY E-MAIL ADDRESS IN FUTURE WILL (ONLY) BE:
*** [log in to unmask]
*** PLEASE CHANGE YOUR ADDRESS BOOK ACCORDINGLY.
*** THANK YOU
Dr. Paul C. Driscoll
Reader in Structural Biology, Joint UCL/LICR NMR Laboratory
---------------------------------------------------------------------
Dr. Paul C. Driscoll |Office (answer)phone: (44)-20 7679 7035
Dept. Biochem. & Mol. Biol. | Department fax: (44)-20 7679 7193
University College London | [log in to unmask]
Gower Street, London WC1E 6BT| http://www.biochem.ucl.ac.uk/~driscoll
|