In this section the people, personalities and debates on the @fedora-devel mailing list are summarized.
Contributing Writer: Oisin Feeley
Fedora 11 Alpha Release Activities
There was a flurry of activity related to the
Fedora 11 Alpha release (scheduled for 2009-02-03). Denis Leroy inquired on 2009-01-21 what had happened to the freeze, originally scheduled for the previous day, and whether all builds in rawhide were queued until after the freeze. Mamoru Tasaka responded with a link to Jesse Keating's explanation that the freeze is a non-blocking freeze which allows targeted fixes to be made. Tom Lane wanted an "all-clear signal that the alpha tag has been made and we can go back to breaking rawhide ;-)" Jesse created  the
alpha tag and apologized for slacking on it. He suggested that if many dependencies were going to be broken by Tom's
mysql-5.1 push that Tom should ask for a
koji tag specifically to land it and build all the deps for it before moving it into
rawhide itself.Josh Boyer demonstrated how the
Koji command-line can be used to answer queries about what tags are present:
$koji list-tags -- grep f11-alpha $koji list-tag-inheritance f11-alpha
Rahul Sundaram requested that knowledgeable folks would help build the Release Notes for
Fedora 11 by adding relevant information to the wiki. After Rahul got the ball rolling, with some information on the use of
ext4 as the default filesystem, the experimental provision of the
btrfs filesystem and more, Richard W.M. Jones added information on the
MinGW windows cross-compiler and Todd Zullinger added information about
The 2009-01-23 Rawhide Report contained some large lists of broken dependencies which were pounced on by the respective developers. As the majority were due to the new
MySQL mentioned above Jesse Keating asked why his advice to use a special tag had been ignored. Tom Lane replied that there had been no objections when he mooted the idea a week ago and that a non-standard tag would cause more work for affected developers than the current rebuilds. Jesse re-iterated his request to "[p]lease consider using it in the future if you're going to break such a wide array of packages."
Richard W.M. Jones reported problems using
Rawhide. Tom London suggested and Richard W.M. Jones confirmed that reverting to
sqlite-3.6.7-1.fc11.x86.64 fixed the problems. It transpired that there was indeed an
SQLite bug which was quickly fixed by Panu Matilainen.
Minimalist Root Login to X ?
Warren Togami suggested "mak[ing] root logins from GDM a stripped down desktop with only a terminal and a menu with only configuration tools [and making the desktop] ugly and with a very obvious note explaining why [users] shouldn't be logged in as root."
"Nodata" was among those who wondered if Warren's use cases "[...] where /home filesystem is full and logins fail, or /home is remote and inaccessible[...]" were anything other than odd edge cases. Jeff Spaleta and Chris Adams expanded upon this line of thought: "[...] if /home is full, can users really not log in? If that is the case, that's broke and should be fixed. The user should be able to log in and remove files."
The impetus for this discussion may have been another thread which asserted that the denial of root login via
Fedora 10 systems made it too difficult to maintain said systems. The thread yielded good examples by Jud Craft and Dave Airlie of arguments that such modifications merely penalized experienced users and failed to enhance security as the users could just login as root on the console anyway. As an aside Benjamin LaHaise brought up the issue that
Ctrl+Alt+F2 no longer worked. DanHorák explained that "F2-6 are blocked when you have getty running on vt1 (/etc/event.d/tty1 is the same tty[2-6]) and Xorg server runs on vt1 too (gdm runs with --force-active-vt) Then there are messages like `unable to switch vt' in /var/log/Xorg.log. [Such behavior] requires manual editing of at least /etc/event.d/tty1, it should not happen in default setups." Nicolas Mailhot suggested an imperfect upgrade as another possible cause. A further nugget of information revealed in the thread was as
Fedora 10 had implemented
hiddenmenu as a default in grub it was best to hold down any key once the
BIOS had finished the
POST routine. Jesse Keating suggested the
shift key as it typically had no bindings either in
grub. Andrew Haley pointed out that many of the recent changes were breaking established use patterns. Kevin Kofler and Christopher Wickert suggested that anyone who wished to revert to the previous status should just edit
/etc/pam.d/gdm to comment out
auth required pam_succeed_if.so user != root quiet
Back in the later thread which sought to deal with some of the difficulties raised above Tom `spot' Callaway suggested: "A `Rescue Mode' in GDM which goes to a root session with minimal apps, marked as "Rescue Mode", rather than a root X login (even though it does need root credentials)." Lyos Gemini Norezel preferred that "[...] the root login should use the user selected interface (gnome, kde, xfce, etc)" but Matthew Woehlke emphasized the maintenance benefits of choosing a single Desktop Environment and forcing that as the safe root login.
Variations on this topic have been covered previously in FWN#133 and FWN#103
Fedora Geo Spin for USB Key and LiveCD
Yaakov Nemoy announced a "[...] respin of Fedora with packages for doing OSM and cartography installed out of the box, or included on a LiveCD and/or LiveUSB. For OSM people, the primary advantage is a live usb stick that can be used at mapping parties to save time cono/guring user computers to do mapping. The USB stick can then be brought home, and the user can continue doing mapping there."
 Open Street Mapping http://en.wikipedia.org/wiki/OpenStreetMap
Draft Guidelines for Approving provenpackagers
Jesse Keating drafted a definition of `provenpackager' (see FWN#151[2)]. Alex Lancaster was worried that too many hoops would mean that maintainers such as himself would lose motivation to continue their work.
As a subsidiary concern Alex was worried that there were still some packages not being opened up. KevinKofler assured Alex that he would become a `provenpackager' based up his sterling work and Jesse confirmed that this redefinition and re-seeding of the `provenpackager' group was in part to address such concerns.
Cloning of Bug Reports ?
Jóhann B. Guðmundsson asked for input, in the form of suggestions and votes, as to whether Bug Hunters (which later seemed to mean testers, but not triagers) should file a separate bug entry for each of: past supported release, current release and rawhide or just annotate a bug for one of the former with a note that it was present in the others.
There was general agreement that mailing list votes were ineffective and unwanted.
Kevin Kofler objected to the tack taken by Jóhann which seemed to assume an authority over a decision which would affect not just QA, testing and triage teams but also packagers and maintainers. It appeared that the matter would be elevated to FESCo for a decision but as of going to press this had not happened.
Mark McLoughlin suggested a more flexible policy and warned that "[...] you can be sure you'll have maintainers who haven't read or replied to this thread waking up and getting annoyed that they've 3x bug reports to deal with :-)"
Jesse Keating argued that the multiple bug-entry option was preferable on four heads: 1) that bugs may have different causes in their releases; 2) users of past releases will not be helped by closing bugs on rawhide; 3) bodhi updates are not pushed at the same time; 4) maintainers are the only people with the knowledge to make such a call.