Initializing hardware... slow

[Nicolas Mailhot]

Le samedi 26 novembre 2005 à 13:33 +1100, Bojan Smojver a écrit :
> On Fri, 2005-11-25 at 22:13 +0100, Nicolas Mailhot wrote:
> 
> > Seems related to network calls
> > 
> > ntp startup is sloooow
> > httpd startup -> same
> > gaim startup -> anemic (btw the latest arts is missing lnusertemp, which
> > exposes the well-documented stupidity of having gnome sound depend on
> > kde subsystems)
> 
> How's your DNS? This could all be related to invalid DNS servers
> in /etc/resolv.conf.

Thanks, should have thought of it, the first entry was dead.
Though nscd was running, and some of the side-effects were distinctly
weird (why should dns slow down gnome-term startup when there is a local
cache on the system ?)

> > I'll try bootstart tomorrow - rawhide today is so broken I'm pretty
> > disgusted (gaim/arts/mcop, gdm, selinux, slow startup) If I don't do
> > something else now I'll start writing rude things on the list.
> > 
> > I don't mind a new bug a day but several in a row like this before I've
> > even finished investigating the first one is too much for me. You got to
> > love the destabelizing effect of test 1 - every Red Hat packager seems
> > to have postponed his experimental packages for after test1 was
> > released.
> 
> If things like this make you upset, maybe you shouldn't be running
> Rawhide. It is (unfortunately) not possible to find bugs without
> delivering software to users.

Things are not black and white. I don't mind new bugs. I do mind hitting
bugs/problems reported since before FC4 (making gnome depends on arts)
or discussed on the list in the days before packages are released (the
long thread about people not wanting core fonts mandatory didn't stop
mharris from releasing a modular X that crashed because of core font
stuff) or obvious checks the maintainer skipped before inflicting
breakage on everyone (arts changelog reads "4 nov fix lnusertemp
problem", "14 nov fix lnusertemp", and on the 24th a new version with
broken lnusertemp is released - how difficult would it have been to
check lnusertemp before puching a new package ?)

So yes it's RawHide. Yes it's not supposed to be perfect and some
breakage is acceptable. But no that does not mean packagers don't have
to check their stuff and make sure it's dogfoodable. If only out of
respect to the scores of poor loosers like me that try do do serious
testing and bug reporting. Because I can say too 'it's rawhide, it's
supposed to be broken, I won't spend time reporting anything' and
complain loudly when the stuff I didn't report appears in FC/RHEL.

So you see this cuts both ways. If Red Hat wants free QA and testing Red
Hat has to deliver something that can be tested. Writing "Raw hide will
eat your brainz" is a bit too easy. Easy and convenient, but false.
Especially when a release is entering its testing phase.

And BTW I don't have any axe to grind with the people who made the
mistakes I listed. Everyone makes mistakes now and then. But I won't let
write that because it's RawHide there is no need to acknowledge any
problem, and testers should not be upset when well-known issues are
pushed to rawhide. There are enough new bugs to find without having to
report the same stuff again and again.

-- 
Nicolas Mailhot
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 197 bytes
Desc: Ceci est une partie de message num?riquement sign?e
URL: <http://listman.redhat.com/archives/fedora-devel-list/attachments/20051126/46d52f69/attachment.sig>