octave-dev

[OctDev] octave-forge on solaris 8

[Aaron J. <ja...@ms...>]

I hope this is the proper place for this...

I am trying to build octave-forge on a solaris 8 system.  I have never  
done this before, so I'm not sure what to expect so I have a few  
questions.  My first question, is that the version of octave I'm using  
is 2.1.71 while the version of octave-forge is octave-forge-2005.06.13.  
  Is this a bad idea?  I didn't realize the version mismatch until after  
I installed 2.1.71 and I would rather not redo the 9+ hour build  
(octave and octave-forge) if I don't have to.

Doing a

./configure --prefix=/opt/sfw
gmake -k

Seemed to work, but there were a few ld errors along the way.  I guess  
that's what the -k is for.  However doing a

gmake check

gave me some problems.  I did some searching and found that there were  
some problems with bash specific shell scripts.  The admin/mktests.sh  
and admin/run_forge scripts do need bash not sh to run.  Also,  
admin/mktests.sh calls grep -E.  The -E is not a valid switch for  
/usr/bin/grep for solaris 8.  So either the script needs to call  
/usr/xpg4/bin/grep, or it needs to call egrep.  After fixing that I was  
able to run the check.  There were a lot of failed tests and the check  
eventually errored out (See below.  I can send the fntests.log to  
anybody who is interested.).  Is this normal, because octave builds and  
tests very cleanly on this system?

I then did a

gmake install

This also gave me some errors since the install scripts call grep -q.   
Again, -q is not a valid switch for /usr/bin/grep.

After fixing that, when I started up octave I got the following errors:


error: `dispatch' undefined near line 2 column 1
error: near line 2 of file  
`/opt/sfw/share/octave/2.1.71/site/m/octave-forge/comm//PKG_ADD'
error: source: error sourcing file  
`/opt/sfw/share/octave/2.1.71/site/m/octave-forge/comm//PKG_ADD'

I then took the advice of the install script and uninstalled  
octave-forge.  There are some functions in octave-forge that I need  
access to, but I am having problems installing it.  Has anybody gotten  
this to work on a solaris 8 machine?  Should I make sure that the  
versions match and then try everything from scratch?  Thanks for any  
pointers.

Aaron

jackson@helen:octave-forge-2005.06.13 {29} gmake check
admin/mktests.sh
admin/run_forge octave --norc -q fntests.m
FIXES [tests        2 of       13 files]
warning: hankel: column wins anti-diagonal conflict
                                          ---> passes 8 out of 22 tests
main/audio [tests        2 of       11 files]
                                          ---> success
main/cell [tests        1 of        2 files]
                                          ---> success
main/comm [tests        5 of       60 files]
                                          ---> success
main/control [tests        0 of        1 files]
                                          ---> success
main/econometrics [tests        0 of       25 files]
                                          ---> success
main/fixed [tests        0 of       43 files]
                                          ---> success
main/general [tests        5 of       33 files]
                                          ---> success
main/geometry [tests        0 of        7 files]
                                          ---> success
main/ident [tests        0 of        4 files]
                                          ---> success
main/image [tests       34 of       85 files]
                                          ---> passes 186 out of 242  
tests
main/io [tests        0 of        8 files]
                                          ---> success
main/irsa [tests        0 of       14 files]
                                          ---> success
main/linear-algebra [tests        0 of        3 files]
                                          ---> success
main/miscellaneous [tests        2 of       22 files]
                                          ---> passes 4 out of 15 tests
main/optim [tests        3 of       54 files]
warning: fzero: equal signs at both ends of the interval.
                         Using fsolve('sin',0.75) instead
                                          ---> passes 9 out of 10 tests
main/path [tests        2 of        3 files]
                                          ---> success
main/plot [tests        1 of       34 files]
                                          ---> passes 6 out of 7 tests
main/set [tests        2 of        5 files]
                                          ---> passes 8 out of 10 tests
main/signal [tests       19 of       79 files]
warning: grpdelay: setting group delay to 0 at singularity
warning: grpdelay: setting group delay to 0 at singularity
warning: grpdelay: setting group delay to 0 at singularity
warning: grpdelay: setting group delay to 0 at singularity
                                          ---> passes 72 out of 74 tests
main/sparse [tests        8 of       23 files]
warning: PCG: eigenvalue estimate failed: iteration converged too fast.
                                          ---> passes 62 out of 439 tests
main/specfun [tests        3 of       15 files]
                                          ---> success
main/special-matrix [tests        2 of        6 files]
                                          ---> success
main/splines [tests        3 of       13 files]
                                          ---> success
main/statistics [tests        1 of       24 files]
warning: XXX FIXME XXX Tsq return from princomp fails some tests
warning: XXX FIXME XXX Tsq return from princomp fails some tests
                                          ---> passes 10 out of 11 tests
main/strings [tests        2 of       13 files]
                                          ---> passes 25 out of 27 tests
main/struct [tests        0 of       10 files]
                                          ---> success
main/symbolic [tests        0 of        5 files]
                                          ---> success
main/time [tests        5 of        8 files]
                                          ---> passes 13 out of 31 tests
main/vrml [tests        0 of       33 files]
                                          ---> success
extra/civil [tests        0 of        3 files]
                                          ---> success
extra/graceplot [tests        0 of        3 files]
                                          ---> success
extra/integration [tests        0 of       25 files]
                                          ---> success
extra/linear-algebra [tests        1 of        2 files]
                                          ---> passes 4 out of 5 tests
extra/mapping [tests        0 of        4 files]
                                          ---> success
extra/mex [tests        0 of        2 files]
                                          ---> success
extra/ode [tests        0 of        8 files]
                                          ---> success
extra/pdb [tests        0 of        5 files]
                                          ---> success
extra/symband [tests        0 of        7 files]
                                          ---> success
extra/testfun [tests        4 of        7 files]
                                          ---> passes 85 out of 95 tests
extra/tsa [tests        0 of       50 files]
                                          ---> success
passes 742 out of 1238 tests
see fntests.log for details
admin/run_forge octave --norc -q batch_test.m
[main/comm]
 >comms

<< Random Signals Package >>
   Signal Creation:                          PASSED
   Signal Analysis:                          PASSED

<< Source Coding Package >>
   PCM Functions:                            Not tested
   Quantization Functions:                   PASSED

<< Block Coding Package >>
   Cyclic Coding:                            error: `cyclpoly' undefined  
near line 244 column 6

Note: failure expected for octave 2.1.36
[main/fixed]
 >fixed

<< Fixed Point Load Type >>
error: `fixed' undefined near line 152 column 11

Note: failure expected for octave 2.1.36
[main/image]
 >jpgread ... not available
 >jpgwrite ... not available
 >pngread ... not available
 >pngwrite ... not available
[main/struct]
 >getfield
 >setfield
[main/sparse]
error: `nnz' undefined near line 163 column 16
error: evaluating argument list element number 3
error: evaluating if command near line 142, column 1
error: near line 169 of file  
`/export/home/jackson/software/octave-forge-2005.06.13/main/sparse/ 
fem_test.m'
gmake: *** [check] Error 1

Re: [OctDev] octave-forge on solaris 8

[Paul K. <pki...@us...>]

On Jul 1, 2005, at 4:34 PM, Aaron Jackson wrote:

> I hope this is the proper place for this...

Yes.

> I am trying to build octave-forge on a solaris 8 system.  I have never 
> done this before, so I'm not sure what to expect so I have a few 
> questions.  My first question, is that the version of octave I'm using 
> is 2.1.71 while the version of octave-forge is 
> octave-forge-2005.06.13.  Is this a bad idea?  I didn't realize the 
> version mismatch until after I installed 2.1.71 and I would rather not 
> redo the 9+ hour build (octave and octave-forge) if I don't have to.

The octave-forge build skips what can't be built (the -k is done in the 
build).  As far as I'm aware everything in 2005.06.013 should build in 
2.1.71, but I haven't tested this combination.

> Doing a
>
> ./configure --prefix=/opt/sfw
> gmake -k
>
> Seemed to work, but there were a few ld errors along the way.  I guess 
> that's what the -k is for.  However doing a
>
> gmake check
>
> gave me some problems.

>   I did some searching and found that there were some problems with 
> bash specific shell scripts.  The admin/mktests.sh and admin/run_forge 
> scripts do need bash not sh to run.

Are there a lot of changes required for it to run on sh?  I'm happy not 
to depend on bash if it is easy enough.  Please send a patch.

> Also, admin/mktests.sh calls grep -E.  The -E is not a valid switch 
> for /usr/bin/grep for solaris 8.  So either the script needs to call 
> /usr/xpg4/bin/grep, or it needs to call egrep.  After fixing that I 
> was able to run the check.

egrep sounds like a better option.  However, I don't see anything in 
the pattern which requires extended expressions.  Will it work with 
plain grep and no -E option?


>   There were a lot of failed tests and the check eventually errored 
> out (See below.  I can send the fntests.log to anybody who is 
> interested.).  Is this normal, because octave builds and tests very 
> cleanly on this system?

Most tests should work.   From below I suspect that for some reason you 
cannot use compiled functions in oct files.   Does your octave/src 
directory contain *.oct files?  Can you run them from the octave 
command line?

If the answer is no, then try

	./configure --enable-shared --disable-static

and rebuild.  At some point this became the default but I don't 
remember when.  You can also scan through config.log to see if shared 
is enabled if you want to save yourself a nine hour test.

>
> I then did a
>
> gmake install
>
> This also gave me some errors since the install scripts call grep -q.  
> Again, -q is not a valid switch for /usr/bin/grep.

Easy enough to fix these with a >/dev/null.  Might break mingw install 
though.

>
> After fixing that, when I started up octave I got the following errors:
>
>
> error: `dispatch' undefined near line 2 column 1
> error: near line 2 of file 
> `/opt/sfw/share/octave/2.1.71/site/m/octave-forge/comm//PKG_ADD'
> error: source: error sourcing file 
> `/opt/sfw/share/octave/2.1.71/site/m/octave-forge/comm//PKG_ADD'
>
> I then took the advice of the install script and uninstalled 
> octave-forge.  There are some functions in octave-forge that I need 
> access to, but I am having problems installing it.  Has anybody gotten 
> this to work on a solaris 8 machine?  Should I make sure that the 
> versions match and then try everything from scratch?  Thanks for any 
> pointers.

Check the mailing list archives at http://www.octave.org.  There are 
occasional posts Re: Solaris.

- Paul

Re: [OctDev] octave-forge on solaris 8

[Dmitri A. S. <das...@gm...>]

FWIW:

I would recommend to install gnu version even of those utils
which are provided with Solaris 8. perl, sed, and, as you found, grep
are some of the examples.
Also, Solaris' /bin/sh is not POSIX compliant, so you may want to
replace
SHELL=/bin/sh
with
SHELL=/usr/local/bin/bash
in a couple makefile/makeconf scripts

You can get most of the stuff at www.sunfreeware.com

Sincerely,

Dmitri.
-- 

Re: [OctDev] octave-forge on solaris 8

[Aaron J. <ja...@ms...>]

On Jul 1, 2005, at 10:55 PM, Dmitri A. Sergatskov wrote:

> FWIW:
>
> I would recommend to install gnu version even of those utils
> which are provided with Solaris 8. perl, sed, and, as you found, grep
> are some of the examples.
> Also, Solaris' /bin/sh is not POSIX compliant, so you may want to
> replace
> SHELL=/bin/sh
> with
> SHELL=/usr/local/bin/bash
> in a couple makefile/makeconf scripts
>
> You can get most of the stuff at www.sunfreeware.com

The regular Solaris commands listed above are not POSIX compliant, 
however there are POSIX compliant versions in /usr/xpg4/bin.  I just 
tested it and it is only necessary to put /usr/xpg4/bin first in the 
PATH variable to run the scripts as is.  As a side effect I see that 
main/image/bestblk.m, main/image/bweuler.m and main/image/conndef.m are 
missing new lines at their end.  I guess the build process for solaris 
should be

setenv PATH /usr/xpg4/bin:$PATH
./configure
gmake -k
gmake check
gmake install

Thanks.

Aaron