Bug#752351: pdl: hardcodes /usr/lib/perl5

[Niko Tyni]

clone 752351 -1
reassign -1 libpdl-stats-perl 0.6.5-1
retitle -1 libpdl-stats-perl: don't run add_doc.pl at build time
severity -1 normal
thanks

On Tue, Jun 24, 2014 at 10:19:52PM +0200, Henning Glawe wrote:
> On Tue, Jun 24, 2014 at 10:41:00PM +0300, Niko Tyni wrote:

> > > my PDL package has a trigger installed for doc/doc-index rebuild if 
> > > any package installs sth into its namespace, so no explicite action on the
> > > side of a dependee pkg should be necessary.
> > > I suspect that, besides PDL being possibly broken by the transition, the
> > > dependee libpdl-stats-perl is doing something explicitely that it
> > > shouldn't...
> > 
> > The PDL-Stats upstream Makefile.PL has this postamble:
> > 
> > install ::
> >         @echo "Updating PDL documentation database ..."
> >         @$(PERL) add_doc.pl
> > 
> > and add_doc.pl uses PDL::Doc methods to update the doc database
> > it finds on @INC.
> > 
> > So should the libpdl-stats-perl Debian package be disabling that and
> > relying on the trigger instead?

> I definitely think so: why should any package at build time do something that
> is related to the system that it may be installed into...
> 
> Background: pdl, as perl's equivalent to numpy/scipy/matlab, comes with an
> integrated documentation system, allowing to search for documentation of all
> the installed packages related to it.
> the documentation index, searched by tools like pdldoc or the interactive
> shells, should always be consistent with what is actually installed on a
> system.

Ah, thanks. I get it now. Cloning a separate bug against libpdl-stats-perl.
It looks like the trigger is working fine, the bug is just about
the PDL-Stats build system uselessly running add_doc.pl.

It should be enough to just patch away the above postamble on Debian systems
AFAICS.
-- 
Niko Tyni   ntyni at debian.org