The tar balls for modified packages in the octave-forge packages. (integration and secs1d and secs2d)

[Tatsuro MATSUOKA]

Hello

As before, I discussed problems of the octave-forge the integration and the 
secs1d and secs2d
packages.
(Discussions are cited at the end of mail.)


The problems have been almost solved by David (integration) and Carlo (secs).  
However, the results
are uploaded in the svn trees.  To save the time of the users, I packaged them 
and uploaded in the
following cite.

http://www.geocities.jp/tmacchant/additional.html
The tar balls for modified packages in the octave-forge packages

This site is basically for the cygwin user, however, these problems are common 
for all platforms.  

It will be grateful for me that they are helpful for the Octave-forge users.

Thanks here David and Carlo for their corporations.

Regards

Tatsuro



*******************
integration
*******************
>> The issue is that the package manager doesn't respect sub-directories in
>> packages, unless you specifically do so in the PKG_ADD and PKG_DEL
>> files. This is done in secs2d, though needs fixing, but wasn't done for
>> the test function in the integration package, as these are testing
>> functions that probably shouldn't be used.. In fact I suspect the only
>> reason you could use run2dtests was that you were in the directory
>> containing it..
>>
>> I see we have three choices.
>>
>> 1) Leave the situation as is, and don't make any of the test code
>> available in normal use.. For those wanting it that can do
>>
>> lst = pkg("list");
>> pth = "":
>> for i = 1 : length (lst)
>>   if (strcmp (lst{i}.name, "integration"))
>>      pth = lst{i}.dir;
>>      break;
>>   endif
>> endfor
>>
>> addpath (fullfile (dir, "test"));
>> addpath (fullfile (dir, "testfun"));
>>
>> to get the test code.
>>
>> 2) Make sure the test code isn't included in the integration package,
>> but is only in the SVN
>>
>> 3) Make the test code be available by using explicit addpath commands in
>> PKG_ADD..
>>
>> What solution is preferred?
>>
>> D.
>>     
>
> I think the third is preffered because the test scripts give the users the 
> coding information and 
give
> them peace of mind.
>   

Ok, I've created a PKG_ADD and PKG_DEL file that adds them to the path
and committed it. Note that the run2dtests seems to fail however for
another reason that I haven't chased down.





******************
secs1d and secs2d
******************
Hi,
I checked in the changes.
On my system both packages seem work.
c.

On 17/dic/07, at 20:31, Tatsuro MATSUOKA wrote:

> Hi Carlo
>
> Thank you for your reply.
> I will wait for your reports.
>
> Regards!
>
> Tatsuro
>
> --- Carlo de Falco <address@hidden> wrote:
>
>>
>> On 17/dic/07, at 11:37, Carlo de Falco wrote:
>>
>>> Hi Tatsuro,
>>>
>>> Thanks for pointing this out.
>>> The problem seems due to a change in the pkg system since I last
>>> checked in the code.
>>> As now oct files are stored in an architecture dependent directory
>>> the path of
>>> m-files relative to them has changed.
>>>
>>> Changing the line:
>>>
>>> dirlist= {"Utilities","DDG","DDN"};
>>>
>>> to
>>>
>>> dirlist= {"../Utilities","../DDG","../DDN"};
>>>
>>> in both PKG_ADD and PKG_DEL should do the trick but:
>>>
>>> 1) I don't know whether this is a clean solution
>>> 2) I cannot check in the change at the moment as I do not have  
>>> SVN on
>>> my machine
>>
>> Oh, and I forgot:
>>
>> 3) If this change is made, the dependency on octave should also be
>> changed
>> to the first version including system dependent oct-files  
>> directories in
>> the DESCRIPTION of the packages
>>
>> I will install svn later today or tomorrow and try to check-in this
>> changes.
>>
>>> As the logic in PKG_ADD and PKG_DEL was proposed by David  
>>> Bateman, he
>>> might want to comment on point 1)
>>>
>>> Best regards,
>>> Carlo


--------------------------------------
New Design Yahoo! JAPAN  2008/01/01
http://pr.mail.yahoo.co.jp/newdesign/