Re: Providing an easy additional way of avoiding uploading non-redistributable packages

[Joerg Sonnenberger <joerg%bec.de@localhost>]

On Sat, May 23, 2020 at 12:56:13AM +0000, John Klos wrote:
> > I disagree with both. Seriously, it is enabled by default. How much
> > easier do you want something to be discoverable? It is also not the only
> > tool we have around for the job, so this is just an excuse. Uploading
> > from a different machine doesn't make sense either, since the script can
> > *still* be used to just do the filtering and prep work. Been there, done
> > that when uploading weekly results when visiting university.
> 
> There seem to be lots of things that pbulk does that we don't necessarily
> want to do all at once. If you're opposed to making things easier for people
> who don't use pbulk, then have you considered making pbulk more modular, so
> people can use functionalities of pbulk without needing a full chroot?

We already have a tool in lintpkgsrc that does various things of those
outside the scope of bulk build. That doesn't need to break the build of
packages either.

> I'd love to be able to:
> 
> * list packages which have NO_BIN_ON_FTP, to make an exclude list (for rsync)

lintpkgsrc -R.

> * list packages and dependencies which are out of date (given a pkgsrc tree
> and a binary repository)

This isn't even something that pbulk does in isolation.

> * make pkg_summary.gz (& .bz2) and SHA512.bz2

"pkg_info -X" and "digest sha512" is literally all there is to it. The
only difference from the "dumb" invocation is that pbulk matches the
list of files it does upload.

Joerg