On 2018/07/11 10:15, Raf Czlonka wrote:
> On Wed, Jul 11, 2018 at 09:24:56AM BST, Stuart Henderson wrote:
> > On 2018/07/07 23:00, Raf Czlonka wrote:
> > > On Sat, Jul 07, 2018 at 12:40:47AM BST, Stuart Henderson wrote:
> > > > [...]
> > > >
> > > > +MASTER_SITES = http://www.spinnaker.de/debian/
> >
> > Oh this is a redirect to https now, might as well change that and avoid
> > the extra roundtrip.
>
> Shouldn't we change it to https://www.spinnaker.de/lbdb/download/ ?
>
> debian -> lbdb feels more "intuitive", for a lack of a better word.
yep.
> > > > +BUILD_DEPENDS = ${RUN_DEPENDS-main}
> > > > +LIB_DEPENDS-main = converters/libiconv
> > > > +RUN_DEPENDS-main = mail/abook
> > >
> > > There's not need for abook to be a run dependency - m_abook module
> > > is only usable if it is both specified in METHODS *and* abook
> > > executable present, ignored otherwise.
> >
> > It's a really small dependency though, and the only extra thing it pulls
> > in is gettext which mutt users will have anyway, so I don't see a lot of
> > point
> > in /not/ including it, it does make things more consistent.
>
> Sure - however, I was thinking ahead.
>
> I.e., the latest version of lbdb has goobook[0] and khard[1] modules.
> Personally, I don't think each and every program which *enhances*
> lbdb, should be a hard dependency. Even in Debian all of these are
> only *suggested* packages[2]. Personally, I don't even think that
> the -ldap flavor is necessary - a pkg-readme with all the optional
> packages, i.e.:
>
> - Net::LDAP
> - abook
> - goobook
> - khard, which would pull vdirsyncer
> - ...
>
> would suffice, IMO.
>
> This would:
>
> - remove the need for hard dependencies so no extra software, which
> otherwise would potentially not have been used, needs to be
> installed
>
> - if LDAP support is "merged" to the main flavor, there's one less
> flavor to maintain
>
> - any additional *enhancement* are added to the readme
>
> It somehow feels simpler and easier.
>
> Otherwise we'll end up with two packages, where they pull 3-4
> additional programs and the only difference is that one pull an an
> extra Perl module.
>
> Don't some ports already do just that?
>
> Or we go the other way and each and we add each and every of these as
> hard dependencies but the chain grows...
>
> Anyway, food for thought :^)
-ldap is a multi package rather than a flavour, it contains only the
additional files produced when ldap is enabled rather than being an
"either/or" thing.
The usual procedure with "modular" ports using multi packages is to
include things with lightweight/no extra dependencies directly in the
main package, and place things with heavier dependencies in a subpackage.
> [0] https://pypi.org/project/goobook/
> [1] https://github.com/scheibler/khard
> [2] https://packages.debian.org/sid/lbdb
I think these would be good candidates for additional multi packages.
While we *could* merge them into the main package and add a README telling
people how to actually use them, it's a lot more convenient for the user
to just run pkg_add once.
It's a pain to have a proliferation of flavours as they have to be tested
separately. Multi packages are simple in this respect, there's only a single
build run, it produces the same resulting set of packages every time, and
if you're working from ports and want to install the whole lot you can just
do "make install-all".
>
> > > Now, the main issue. The m_abook module doesn't work properly due
> > > to tab escape sequences (\t) being used in sed regexes - something
> > > sed(1) in base does not cope well with.
> > >
> > > All that module does is running the below command:
> > >
> > > abook --datafile $HOME/.abook/addressbook --mutt-query "$@" \
> > > | sed -e '1d;s/\([^\t]*\t[^\t]*\).*/\1\tabook/'
> > >
> > > but, due to the above, based sed(1) cuts the result short on 2nd
> > > 't' in the string, i.e.:
> > >
> > > firstname.surn...@example.org Firstabook
> > >
> > > or, if the string contains only one 't' or none at all:
> > >
> > > first.n...@example.org My Name tabook
> > >
> > > It works fine with gsed, which was obviously used when that module
> > > was written/tested, but then gsed would need to become a run
> > > dependency - something I am not keen on.
> > >
> > > The "easiest" thing to do, would be to patch /usr/local/lib/lbdb/m_abook
> > > and use literal tabs:
> > >
> > > - | sed -e '1d;s/\([^\t]*\t[^\t]*\).*/\1\tabook/'
> > > + | sed -e '1d;s/\([^ ]* [^ ]*\).*/\1 abook/'
> > >
> > > After the above change, it all seems to be working fine.
> > >
> > > Thanks for looking into it.
> > >
> > > Raf
> >
> > Given the existing mess in m_muttalias.sh.in I think that would be ok :)
>
> That's what I thought :^)
>
> Thanks,
>
> Raf
>
