control: tags -1 + pending

On Mon, Oct 23, 2017 at 9:38 PM, Guido Günther <a...@sigxcpu.org> wrote:
> Hi,
> On Sun, Oct 22, 2017 at 12:42:29PM +0200, Michael Stapelberg wrote:
>> On Sun, Oct 22, 2017 at 12:28 PM, Guido Günther <a...@sigxcpu.org> wrote:
>> > Hi,
>> > On Sun, Oct 22, 2017 at 11:15:20AM +0200, Michael Stapelberg wrote:
>> >> Thanks for filing this.
>> >>
>> >> On Sun, Oct 22, 2017 at 9:21 AM, Guido Günther <a...@sigxcpu.org> wrote:
>> >> > Package: pk4
>> >> > Version: 2
>> >> > Severity: wishlist
>> >> >
>> >> > Hi,
>> >> > it would be nice if pk4 would allow to use "gbp import-dsc" to unpack
>> >> > the donwloaded sources in downloadDSCAndUnpack so users:
>> >> >
>> >> >    - get a git archive right away
>> >> >    - can reuise their gbp configuration such as the configured builder,
>> >> >    - pristine-tar, ... right away
>> >>
>> >> I think we could introduce a hook which would replace the default
>> >> behavior of dpkg-source -x, taking as arguments the path to the DSC
>> >> and the destination directory.
>> >
>> > Sounds good.
>> >
>> >>
>> >> I wonder whether the gbp hook should live in the git-buildpackage
>> >> Debian package, though? That way, you could maintain it directly. If
>> >
>> > Sure. It would be nice if setup for users would the be as simple as
>> >
>> >   ln -s /usr/share/doc/git-buildpackage/examples/pk4 ~/pk4/pk4.deb822.d/gbp
>>
>> Almost:
>>
>> mkdir -p ~/.config/pk4/hooks-enabled/unpack/
>> ln -s /usr/share/pk4/hooks-available/unpack/gbp \
>> ~/.config/pk4/hooks-enabled/unpack/
>>
>> Regarding the symlink target, could we ship
>> /usr/share/pk4/hooks-available/unpack/gbp in git-buildpackage? That
>> way, all hooks would be in the same directory. This is similar to how
>> shell tab completion files are shipped.
>
> Shipped now with the next gbp version:
>
>     https://github.com/agx/git-buildpackage/blob/master/debian/pk4

Neat!

I just committed
https://github.com/Debian/pk4/commit/797dc0b887abbc482a7a095d687b710509a80816,
upload to Debian follows in a second.

One thing I noticed: the resulting branches are master, pk4 and
upstream, which the currently checked out branch being master.
Shouldn’t the only two branches be pk4 and upstream?

>
>> >> you think that’s not a good idea, could you suggest how the hook
>> >> should be implemented? I’m envisioning something like this (untested):
>> >>
>> >> #!/bin/sh
>> >> set -e
>> >> mkdir -p "$2"
>> >> cd "$2"
>> >> git init
>> >> gbp import-dsc "$1"
>> >
>> > #!/bin/sh
>> > set -e
>> > gbp import-dsc "$1" "$2"
>> >
>> > is enough (gbp will do the rest). That way we could also support
>> > incremental imports (that is if the directory is already there we simply
>> > import the new version on top of it so the use can diff between the old
>> > an new version.
>>
>> Note that the pk4 output directories contain the version number, so I
>> think incremental imports wouldn’t work well.
>>
>> >
>> >> Side note: I think we should be fairly clear about the difference
>> >> between a gbp-from-dsc repo and a gbp-from-gbp-clone repo, to not
>> >> confuse our users.
>> >
>> > Yeah. I was thinking of putting a .git/gbp.conf into the repo that sets
>> >
>> >   [DEFAULT]
>> >   upstream-branch = upstream
>> >   debian-branch = pk4
>> >
>> > This would
>> >
>> >  - make sure we override settings any branch settings in debian/gbp.conf
>> >    which we don't care about (since we're not cloning from Vcs-Git:
>> >  - Having the default branch named pk4 would make it obvious that this
>> >    is s.th. special.
>> >
>> > What do you think? In this case it would rather be more like your script
>> > above:
>>
>> Sounds good to me.
>>
>> >
>> > #!/bin/sh
>> > set -e
>> >
>> > if [ ! -d $2 ]; then
>>
>> nit: use "$2" here as well
>>
>> >
>> >   git init "$2"
>> >
>> >   cat <<EOF > "$2"/.git/gbp.conf
>>
>> I suggest to add as a comment what you wrote above for the benefit of
>> readers of the hook :).
>>
>> >   [DEFAULT]
>> >   upstream-branch = upstream
>> >   debian-branch = pk4
>> >   EOF
>> > fi
>> >
>> > cd "$2"
>> > gbp import-dsc "$1"
>> >
>> >
>> > Does this sound reasonable? I would then also provide a script that can
>> > be used with pk4-replace.
>>
>> I don’t quite follow. What sort of script is required for that?
>
> Probably not even a script but a post-build hook that cats the name of
> the generated changes file to /proc/self/fd/3.

Ah, I see.

> Cheers,
>  -- Guido



-- 
Best regards,
Michael

Reply via email to