Re: Issues with build Fedora RPMs of octave packages

[David Bateman]

Michael Goffioul wrote:
> On 9/21/07, David Bateman <address@hidden> wrote:
>> Michael Goffioul wrote:
>>> I guess this will completely defeat the current strategy used in the java
>>> package to find octave.jar at runtime... :-(
>>> I think I'll then install octave.jar also in libexec, even if it's
>>> arch-independent.
>>>
>> Michael,
>>
>> I haven't applied this and won't till at least one distribution packager
>> confirms that it is sufficient to meet their packaging rules and
>> acceptable. I'd be happy if someone proposes an alternative location.
>> The patch I propose only needs a single line changed to propose any
>> alternative location.
>>
>> I also wonder what should be done on Windows.. Should the architecture
>> dependent files just be grouped with the package files as currently
>> done, or for consistency should they be in
>>
>> <libexec>/octave/packages?getarch()/<package>
>>
>> as the proposed patch does. I'd be happy either way, and the change to
>> the previous patch to allow the previous behavior under windows is minimal..
> 
> I'd go for the same location on all platforms. However, my first
> reaction was wrong. 

That's the easiest for me..

> I forgot that the look for octave.jar is not based on some
> Win32 specific API (GetModuleFileName), but on the octave load_path
> class. So, I should be able to put the octave.jar file in any directory as
> long as it's in octave's path. I'll just have to adapt the code of __java__.cc
> a little bit (after you commit your changes, if you do).
> 

Ok, I'll wait for feedback from Orion.. In any case there was a trivial
bug in the code I sent, but it only shows up when removing user
installed packages, so its not an issue for this discussion.

D.