[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Broken .octave_hist
From: |
Joe Koski |
Subject: |
Re: Broken .octave_hist |
Date: |
Sat, 06 Dec 2003 16:21:24 -0700 |
User-agent: |
Microsoft-Entourage/10.1.1.2418 |
on 12/6/03 2:57 PM, Per Persson at address@hidden wrote:
>
> On Saturday, December 6, 2003, at 09:40 PM, Joe Koski wrote:
>
>> As a new octave 2.1.50 user on Mac OS X 10.2.8, I installed octave in
>> usr/local from binaries (not from Fink).
>
> Could you be more specific as to where you got the binaries and if
> possible how they were built? (Related to dispatch error below)
Thanks for the quick response.
My binaries came from http://hpc.sourceforge.net/. They appear to be built
with the gcc-3.3/g77-3.4 combination mentioned on that site. The install is
a simple "sudo tar -xvf octave.tar -C /", and octave (and history) worked
well for over a week until I discovered the need for the spline routines.
(Incidently, I had tried building octave 2.1.50 from source, but that plan
was abandoned when I couldn't get the OSF readline "make" to work because of
missing library files:
ld: Undefined symbols:
restFP
saveFP
_tgoto
_tputs
_tgetent
_tgetflag
_tgetnum
_tgetstr
/usr/bin/libtool: internal link edit command failed
make[1]: *** [libreadline.4.3.dylib] Error 1
make: [shared] Error 2 (ignored)
At that point I gave up and installed the binaries. Gnuplot 3.8j.0 did build
successfully.)
>
>> Everything worked fine until I
>> needed some cubic spline routines, and installed (from source)
>> octave-forge-2003.06.02. The spline routines from octave-forge seem to
>> be
>> working fine.
>
> That's good;-)
>
>> The building of the octave-forge routines was not smooth, but
>> the "make install" seemed to go OK.
>
> It would help if you elaborated on "not smooth".
The ./configure gave a long list of NOINSTALLS, but I went ahead anyway. The
"make" file exited complaining
/usr/bin/texi2dvi: tex: command not found
/usr/bin/texi2dvi: tex exited with bad status, quitting.
/usr/bin/texi2dvi: see comms.log for errors.
make[2]: *** [comms.dvi] Error 1
rm comms.texi
make[1]: *** [comm/] Error 2
make: *** [main/] Error 2
I don't have tex (as in LaTex?) on my machine and I thought it was related
only to help file construction, so I went ahead with "sudo make install"
anyway. I could send a copy of the .log file if it would help.
I did notice that the octave make file commented on MACOSX during the
install, so it was trying to accommodate OS X.
>
>> (As an aside, I wish that we could fix
>> more OSF make files so that they will work properly with OS X and
>> limit the
>> need for Fink, etc. to exist.)
>
> AFAIK, the instructions in octave-forges INSTALL.MacOSX file (below)
> works.
> -----
> Add NOINSTALL to extra/linear-algebra ('touch
> extra/linear-algebra/NOINSTALL')
> Add NOINSTALL to main/sparse ('touch main/sparse/NOINSTALL')
>
> Set PROGS to an empty string (i.e. 'PROGS =') in FIXES/Makefile (Can't
> override builtins on OSX, yet)
>
> Run './configure' (run './autogen.sh' first if you are working with CVS
> sources)
> Run 'make' and 'sudo make install'
> -----
> The three additional steps before configure are not that hard and,
> unfortunately, necessary until someone volunteers to fix that code.
>
>>
>> About the time of the octave-forge installation, my history of
>> commands in
>> .octave_hist stopped being updated.
>
> No idea. What is the value of variables 'history_file' and
> 'saving_history'?
>
> octave:1> history_file
> history_file = /Users/per/.octave_hist
> octave:2> saving_history
> saving_history = 1
>
In checking this, it appears that my who -long and whos may be broken,
because they donĀ¹t list anything, but my variables are
octave:2> whos
octave:3> who -long
octave:4> saving_history
saving_history = 1
octave:5> history_file
history_file = /Users/joekoski/.octave_hist
>
>> Also, now when I start octave, I get the
>> error
>>
>> error: `dispatch' undefined near line 2 column 1
>> error: near line 2 of file
>> `/usr/local/share/octave/2.1.50/site/m/octave-forge/comm//PKG_ADD'
>
> Dispatch seems to work fine here.
> OS X 10.2.8, octave 2.1.50 built from source.
> gcc version 3.3 20030304 (Apple Computer, Inc. build 1435)
>
> Was the octave binary built with support for dynamic loading of
> compiled functions (.oct files)?
> This is a little hard to find out on OS X since
> 'octave_config_info("dld")' returns 0 even if octave does support .oct
> files (Ugh!).
> What does octave_config_info("MKOCTFILE_SH_LDFLAGS") give you?
octave:8> octave_config_info("MKOCTFILE_SH_LDFLAGS")
ans = -bundle -bundle_loader /usr/local/bin/octave-2.1.50
>
> Is the dir
> /usr/local/libexec/octave/2.1.50/site/oct/powerpc-apple-darwin6.6/
> octave-forge populated with .oct files (including dispatch.oct)?
[pcp01876656pcs:oct/powerpc-apple-darwin6.8/octave-forge] joekoski% ls
SBBacksub.oct bchdeco.oct gf.oct greshape.oct mex.o
SBEig.oct bchenco.oct gfilter.oct gsqrt.oct pretty.oct
SBFactor.oct chol.oct ginv.oct gsum.oct primpoly.oct
SBProd.oct cyclgen.oct ginverse.oct gsumsq.oct rand.oct
SBSolve.oct cyclpoly.oct glog.oct isgalois.oct randn.oct
SymBand.oct gdet.oct glu.oct isprimitive.oct rsdec.oct
_errcore.oct gdiag.oct gprod.oct matrix.h rsenc.oct
_gfweight.oct gexp.oct grank.oct mex.h
syndtable.oct
Yes populated (darwin6.8 on my machine), but no dispatch!
>
> Finally, the output of 'octave_config_info' could give some clues to
> whether the binary install is sane.
Attached to this e-mail in case it's helpful. I'll also see if something
obvious jumps out from looking at it.
Sorry for these problems. Otherwise my experience with octave has been quite
good. I got some Matlab data analysis routines that I needed working with
only minor rewrites (including splines). My Octave manual is due to arrive
on Monday, so I've been relying on the web for help.
Joe Koski
>
> /Per
>
octave_config_info.txt
Description: application/applefile
octave_config_info.txt
Description: Text document