Re: [PATCH 3] gnunet-svn (was Re: [PATCH] gnunet.scm -> various changes (description update, adds gnunet-svn, gnunet-gtk-svn, gnurl))

[Jookia]

On Mon, Mar 28, 2016 at 07:49:48PM +0200, Nils Gillmann wrote:
> >> +;; We provide SVN in addition to the older 0.10.1 version,
> >> +;; protocol and API compatibility will be broken when GNUnet bug #4165
> >> +;; is fixed.
> >> +(define-public gnunet-svn
> >> +  (package
> >> +    (name "gnunet-svn")
> >> +    (version "0.10.1.svn36926")
> >> +    (source
> >> +     (origin
> >> +       (method svn-fetch)
> >> +       (uri (svn-reference
> >> +             (url "https://gnunet.org/svn/gnunet/";)
> >> +             (revision 36926)))
> >
> > Are SVN identifiers really that short, or is there a longer version that
> > we should use, as for Git?
>
> I have no idea, I thought this was what we used for svn.
> Most of the -svn is done by Jookia, i just fixed the way they
> were arranged, named, described and some other fixes.

Apologies for not being on IRC, I'm currently away from my main machine so I"m
left with a decaying Arch Linux install that doesn't have an OpenVPN container.

Yes, SVN identifiers are really that short- they're the count of commits done to
the centralized repository. The reason this isn't a thing in Git is because of
decentralization, there'd be no way to keep a monotonic revision number with the
amount of branches and private repositories.

> >> +    (arguments
> >> +     '(#:configure-flags
> >> +       (list (string-append "--with-nssdir=" %output "/lib"))
> >
> > It needs help referring to its own output?
>
> Refer to Jookia, will ping Jookia later about this thread.
>
> >> +       #:parallel-tests? #f
> >> +       ;; test_gnunet_service_arm fails; reported upstream
> >> +       #:tests? #f
> >
> > Okay, is it easy to disable just the failing test? Also, can you include
> > a link to the upstream bug report in this comment?
>
> I don't know which bugreport this refers to, waiting for comment
> by Jookia.

So both of these flags are present in the original gnunet package. Perhaps it'd
be worth seeing if we can remove them in the case they're fied upstream.

> Why? gnunet-$version is the stable release, gnunet-svn is the svn
> release. Both should carry information visible to users on why
> there are different packages.

I'd have to agree with keeping this information, especially since the response
to many GNUnet problems is to "use the SVN version.", which is exactly what
people will find here. It's also useful Guix runs on foreign distributions so
people won't have to spend headaches getting it compiled.

Overall good patches, I'm glad there's a lot of interest and discussion here.
Nothing seems off to me, but those are famous last words and I haven't reviewed
the split patches yet, I'll put it on my 'I'll review it this week' list and
come back next month sometime with replies if it's not merged.

> -- 
> ng
> personal contact: http://krosos.sdf.org
> EDN: https://wiki.c3d2.de/EDN

Jookia.