Re: [gentoo-dev] On the usefulness of eclass changelog

[Alexis Ballier]

On Wed, 31 Oct 2012 12:26:14 -0400
Ian Stakenvicius <axs@g.o> wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA256
>
> On 31/10/12 12:15 PM, Samuli Suominen wrote:
> > On 31/10/12 17:39, Alexis Ballier wrote:
> >> On Wed, 31 Oct 2012 11:35:41 -0400 Ian Stakenvicius
> >> <axs@g.o> wrote:
> >>
> >>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256
> >>>
> >>> On 31/10/12 11:26 AM, Alexis Ballier wrote:
> >>>> On Tue, 30 Oct 2012 18:39:44 -0600 Ryan Hill
> >>>> <dirtyepic@g.o> wrote: [...]
> >>>>>> The file is pointless if not everyone is using it. I've
> >>>>>> offered to remove the file before, and I'm reoffering to
> >>>>>> do so now.
> >>>>>
> >>>>> It's pointy enough for most uses.  Let's keep it that way.
> >>>>
> >>>> I would like to know what are those uses. Here are my
> >>>> thoughts about changelogs:
> >>>>
> >>>> We have cvs logs, cvsweb, etc. So what is the value added
> >>>> from changelogs? Well, those logs are per-file as far as I
> >>>> know, and since a new version of a package means a new
> >>>> .ebuild file, keeping track of changes to packages is painful
> >>>> without a changelog which is global to the whole package.
> >>>> Even if we have all the needed information in the cvs log,
> >>>> changelogs for packages are definitely useful. Now for
> >>>> eclasses the situation is different: I want to know what has
> >>>> recently changed in foo.eclass, what is the fastest way?
> >>>> Search through a changelog file with dozens of absolutely
> >>>> unrelated information, or run cvs log/go to
> >>>> sources.gentoo.org ? I tend to do the latter and find eclass
> >>>> changelogs completely useless.
> >>>>
> >>>
> >>> Cool, you do, that's great.  This doesn't mean others don't use
> >>> a different process tho, and since it *IS* there and is
> >>> *SUPPOSED* to be filled, and it really doesn't hurt to run
> >>> 'echangelog "${msg}" && cvs ci -m "${msg}"' , why not do it?
> >>
> >> so that others are not encouraged to work sub-optimally :)
> >>
> >
> > eclass/ handling should go to repoman and the automated ChangeLog
> > process, should be rather straight forward for knowing person.
> >
>
> I agree, that'd make the whole thing easier.  But until repoman can
> commit in eclass/ it shouldn't be that hard to just run echangelog ,
> as "inefficient" as that may be. :)

Don't get me wrong: thats not running echangelog that is inefficient,
trying to get information from the changelog is. A per-eclass changelog
would be much more useful, as, atm, you'd be able to access the
information without internet connection.

I have yet to see a case where a global eclass changelog is more
efficient and/or convenient.