[Pkg-octave-devel] Packages with PKG_ADD fail to build

Thomas Weber thomas.weber.mail at gmail.com
Mon Mar 31 07:52:50 UTC 2008


Am Sonntag, den 30.03.2008, 23:14 +0200 schrieb Rafael Laboissiere:
> * Thomas Weber <thomas.weber.mail at gmail.com> [2008-03-30 22:53]:
> 
> > On 30/03/08 19:12 +0200, Rafael Laboissiere wrote:
> > > I think that this happens during the initialization of the load path.  Since
> > > "." comes first in the path, Octave tries to load the PKG_ADD file that it
> > > finds there.  Unfortunately, the other components are not yet in the path at
> > > this time, and Octave fails because functions like fullfile are not found.
> > 
> > I wonder if this is a bug or not? The current behavious makes it
> > impossible to have any sensible PKG_ADD file in the startup directory.
> 
> It is a probably a bug, but it should happen so rarely that I doubt the
> Octave developers will be willing to "fix" it.  

I disagree with the "happens rarely". I mean, how do you develop a
package? You work in its directory and install from there all the time
-- which made me adding the "install from directory part". Do people
really tar their code up each time they change something in order to
test their changes?


> It is happening in our case
> because we are running pkg from the unpacked directory.  We might try to
> approach the octave-forge developers and ask them to move the PKG_ADD file
> into a sub-directory.

$ octave --help
  --norc, -f              Don't read any initialization files.

Either the help text for --norc must be changed or PKG_ADD files should
be ignored in the startup directory.


> Another possibility is to rename PKG_ADD just before calling octave and
> renaming after it back to the original name.

The same happens in the check-pkg part. I didn't fix this, because I
think pkg.m should get a "make check" equivalent. There's no sense in
every distribution adding code for this.

	Thomas




More information about the Pkg-octave-devel mailing list