On 2020/09/06 11:26, Edd Barrett wrote:
> Hi,
> 
> On Sun, Sep 06, 2020 at 02:33:12AM +0200, Jeremie Courreges-Anglas wrote:
> > This doesn't catch everything.
> 
> Yeah, I realised shortly after. I think it's because the [^2] part of
> the glob doesn't catch cases where the line ends after `security/gnupg`.
> I spent a little time trying to figure out the right query, but resorted
> to grep.
> 
> Interested to know how ajacoutot@'s script works. Does it use a sqlports
> query?
> 
> > +     security/p5-GPG (no consumer, not updated since import in 2001)
> 
> ^ Shall we kill this one?
> 
> > As Edd notes, the migration path for consumer ports may be as easy as
> > replacing security/gnupg with security/gnupg2 in BUILD, RUN and
> > TEST_DEPENDS, and bump REVISION.
> 
> Below is my diff so far. This includes your latest diff to
> security/gnupg2. An earlier version of this diff in a mini bulk didn't
> break compilation of any of gpg's dependants.
> 
> I wasn't sure if we need to add a pkg spec to the updated RUN/BUILD
> depends, or just updating the pkg path is sufficient.
> 
> `RUN_DEPENDS=gnupg->=2:security/gnupg2` vs. just 
> `RUN_DEPENDS=security/gnupg2`.
> 
> We will also need to add a quirk to ensure that gnupg1 updates to
> gnupg2 properly.

Not needed, the stem is the same so they are both considered as long as
a matching pkgpath is declared.

I think we should just replace security/gnupg with 2.x though.

> > Regarding the user migration path, the compat "gpg2" and "gpgv2"
> > symlinks could be removed for 6.9, or this could be done post-release.
> > I would go with the latter but I don't feel strongly about it.
> 
> I'm thinking post-release too. We are bound to encounter breakage, and
> it'd be good to limit that to -current if possible.

I don't see harm in keeping symlinks for the existing binaries, they may
be used in user configurations and scripts in various places on the system
and I don't think there's a good reason to break these.

Reply via email to