On Thu, 6 Jan 2005 12:09:40 +0000
Alistair Crooks <agc@pkgsrc.org> wrote:
> On Thu, Jan 06, 2005 at 06:53:59AM -0500, D'Arcy J.M. Cain wrote:
> > I checked this (postgresql74-plpython) package and the error seems
> > to be due to Python not found even though the package was just added
> > in the build according to the log.  Perhaps you have a leftover
> > symlink to a non-existent version of Python and pkg_add doesn't
> > overwrite it. Perhaps try deleting the symlink in /usr/pkg/bin if
> > that is the case.
> 
> There is no symlink to python in the sandbox.

I assume that you mean before the pkg_add of python23.  Does pkg_add
create the symlink like building from source does?

> > > databases/py-postgresql    1       darcy@NetBSD.org
> > 
> > Did you get my note about refetching the tar file?  There was an
> > error when it was first released and you may have the wrong file if
> > you picked it up right after the package was updated.
> 
> Yes, I got your mail, thanks. I don't understand this one - we usually
> work around modified distfiles with a DIST_SUBDIR setting.

Hmm.  I couldn't find any notes about this other than the example in
pkgsrc/doc in the file Makefile-example which only describes where to
put the option.  Looking at how it is used it still doesn't seem to
describe this situation.

Basically what happened was that there was an incorrect .tgz in the
distribution directory which was fixed just after I updated the package.
 I immediately did a new make makesum and committed the new distinfo. 
However, it looks like it wasn't fast enough.  Some people had already
downloaded the incorrect file.  Those are the ones that need to delete
the .tgz and refetch.

I can't see how DIST_SUBDIR would fix that.  Really I should have made a
new release but I thought I was fast enough.  Perhaps I still should.  A
3.6.1 release, even with no changes, would fix the bulk builds.

-- 
D'Arcy J.M. Cain <darcy@NetBSD.org>
http://www.NetBSD.org/