[gentoo-dev] Re: Did portage 2.1 change default use flags?

[Duncan]

Peter <pete4abw@×××××××.net> posted
pan.2006.06.13.16.57.04.370327@×××××××.net, excerpted below, on  Tue, 13
Jun 2006 12:57:08 -0400:

> On Tue, 13 Jun 2006 18:08:03 +0200, Bo Ørsted Andresen wrote:
>
>> On Monday 12 June 2006 12:57, Bo Ørsted Andresen wrote:
>>> On Monday 12 June 2006 12:42, Peter wrote:
>>> > All of a sudden, emerge -uD --newuse world is showing dozens of
>>> > ebuild that are replaced due to removed use flags.
>>>
>>> Look at the first section of[:]
>>> http://www.gentoo.org/news/en/gwn/20060116-newsletter.xml
>>
>> As far as I can see this is not mentioned in either this weeks GWN, the
>> portage 2.1 release notes or the 2.1 news page [references]. I'm sure a
>> lot of people running stable don't remember the GWN from January.
>> Shouldn't this be mentioned somewhere now?

Valid point.  An informed Gentoo sysadmin is an effective Gentoo sysadmin.
Now sometimes I think Gentoo users (which are by another name Gentoo
sysadmins) shirk their responsibility to stay informed, but this wouldn't
seem to be one of those cases.  Sure, it was there in January, but
forget the folks who forgot, what about folks who started on Gentoo since
then?  The change should at a minimum be in the release notes.  Additional
coverage would be good, but a responsible admin will certainly read the
release notes for a non-micro upgrade of something as central to system
administration as a package manager, and as this is a non-trivial change,
it should at the very minimum be there.

Or put it this way.  If it was in the release notes and I failed to see
it, I'd consider it my problem (I get to keep the pieces, as  the saying
goes). If things start changing out from under me without notice despite my
reading of such documentation, the problem is with the package and its
documentation, and should be considered a bug.

(FWIW, as a responsible sysadmin, I dealt with the changes when they were
first covered, back in January, so I could rest easy that the changes when
they occurred in the portage I was using wouldn't cause any issues.  See
below.  However, as mentioned, that doesn't do a thing for the poor guy,
responsible or not, who started with Gentoo in February and is still
getting his Gentoo legs under him, only to see all this change without any
warning.)

> And, from a user pov, these changes are difficult to assess. It is not
> obvious what removing mysql, or db, or idn, or gmp might mean,
> especially if the user never put them there in the first place! And, how
> to you explain that openoffice-bin now has -java instead of java? Or,
> why gnupg lost bzip2?

Here I can't agree.  A responsible Gentoo admin will be verifying the USE
flags using --pretend or --ask before every package merge.  As such, (s)he
should be reasonably familiar with them.  Sure, (s)he may not know
specifically what each one does on every package that uses it, but he
should have no more trouble here than with the whole Gentoo concept and
use of USE flags in general.  Portage does a good job of flagging changed
flags in bright yellow.  If an admin isn't familiar with that particular
flag, a quick euse -i <flag> (euse is part of gentoolkit) will reveal both
its purpose, and whether it's a global or local USE flag (and if local,
how common its use is, global can be assumed to be quite commonly used
or it wouldn't be global). From there, it's the bog standard process of
deciding whether you want the flag on or off in make.conf, and dealing
with exceptions as they come up in package.use.  As I said, any
responsible Gentoo sysadmin (that is, Gentoo user by another name) should
be comfortable with this process.  If they aren't, there's the entirely
logical question of why they are using Gentoo in the first place.

> Too many things occurred without explanation.

Agreed.

> What _I_ ended up doing was hacking make.conf and essentially put back
> all the changed -use flags until I could examine this further.

Well, aside from the fact that a responsible sysadmin (well, if he had
been here since January to have read the coverage back then) would have
already verified his USE flags without the benefit of use.default  (I long
ago did a search on all such files in the tree, deleted the ones I knew
weren't going to be part of my profile, backed up the others, and did an
emerge --pretend --newuse to figure out what I needed to fix, then after
fixing them added them to the rsync-exclude list so syncs wouldn't be
bringing them back), what you did was basically what any sane Gentoo admin
would have done.  No big deal.  Dealing with USE flags, both with the
initial merge of the package, and when any change, is simply part of the
job.  IMO, a Gentoo admin unprepared to deal with that part of the job
should be asking himself serious questions about why he's using Gentoo in
the first place, and if another distribution wouldn't be better suited to
his wanting the distribution to make those decisions for him.  There are
certainly many distributions out there willing to do so, but part of
Gentoo's distinctness is that it places this power AND responsibility in
the hands of the individual Gentoo sysadmin (aka Gentoo user).  Someone
who doesn't want that... IMO shouldn't be using Gentoo, since that's part
of what /defines/ Gentoo.

> Maybe this corrected an error from prior ebuilds or portage versions.
> But, from where I sit, the cure seems worse than the original problem.

How so?  It's making Gentoo more Gentoo-like.  Taking a decision that
/was/ being made and changed arbitrarily based on what was merged, by the
distribution, and putting that decision back squarely in the hands of the
folks who have, by making the Gentoo choice in the first place, signified
that they WANT the choice of making that decision, AND the responsibility
for doing so.

As I've said, this absolutely should be in the release notes, and
preferably should be in other coverage of the  portage 2.1 changes as
well.  There is IMO no excuse for it not being there.  However, also
IMO, it shouldn't be a problem for any responsible Gentoo sysadmin, other
than asking the very reasonable question of why the change isn't covered
in the documentation.  Other than that, it's simply doing the bog-standard
coping with routine USE flag changes, only there's a few more of them to
deal with than "routine" in this case.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

-- 
gentoo-dev@g.o mailing list