Developments
In this section the people, personalities and debates on the @fedora-devel mailing list are summarized.
Contributing Writer: Oisin Feeley
Fedora 11: OSS and PulseAudio Conflict Resolved by CUSE ?
A thread[1] from November led Warren Togami to suggest[2] a plan to use CUSE[3] as part of a strategy to deprecate the near obsolete Open Sound System (OSS) which wreaks havoc with PulseAudio enabled boxes. The plan included a fallback to OSS for users who really wanted it.
Bastien Nocera was[4] skeptical that CUSE would be ready in time for Fedora 11 and suggested instead that a list of applications using OSS be created so that they could be fixed.
[1] https://www.redhat.com/archives/fedora-devel-list/2008-November/msg01005.html
[2] https://www.redhat.com/archives/fedora-devel-list/2008-November/msg02195.html
[3] Character Devices in User space: http://lwn.net/Articles/308445/
[4] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00872.html
Rawhide Report 2008-12-08
When the latest Rawhide Report logged[1] one maintainers use of cvs-import.sh Dominik Mierzejewski criticised[2] the use of the script for updating. Richard Jones asked[3]: "[I]s this stuff really documented anywhere? I have tended to learn it by osmosis, deduction and reading the horribly complicated rules in Makefile.common."
Jason Tibbitts argued[4] that using cvs-import.sh nullified the potential advantages of using an SCM as it sequestered the sources elsewhere. Jesse Keating disagreed[5] due to ease of use issues.
A direct answer was provided[6] by Patrice Dumas with links to the relevant portions of the wiki.
[1] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00671.html
[2] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00677.html
[3] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00691.html
[4] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00694.html
[5] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00695.html
[6] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00701.html
The D-Bus Problem
Ian Amess asked[1] for the current status of a problem caused by a substantial update of the D-Bus package. The update had resulted in the incapacitation of many packages. The most important of these was PackageKit, the default graphical application for managing software.
Colin Walters decided[2] that reverting the update was necessary and that changes to D-Bus policy would be postponed. PackageKit, and its GNOME and KDE clients were updated[3] by Richard Hughes in an attempt to accommodate the changes. Richard testified that "[o]ver the last two days we've all been working really hard on fixing up all the projects after the DBus update. I know personally I'm closing a duplicate bugzilla every 30 minutes." He noted that the delay between creating an update and pushing it to a mirror was a limiting factor in being able to implement these fixes.
A post to @fedora-announce by Paul Frields explained[4] the series of steps which allowed users to re-enable normal system updates using PackageKit. As of 2008-12-15 this notice also appears at the top of all the Fedora Project wiki pages.
[1] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg01391.html
[2] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg01412.html
[3] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00746.html
[4] https://www.redhat.com/archives/fedora-announce-list/2008-December/msg00012.html
Fedora Com System ?
An exploration of possible ways to alert users to critical information was initiated[1] by Arthur Pemberton. Most ideas seemed to center around some sort of RSS feed enabled by default on the desktop.
[1] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg01347.html
YUM: Enable --skip-broken by Default ?
Aliasing yum update to yum --skip-broken update was suggested[1] by Steven Moix as a way to prevent a lot of recurring support problems by eliminating dependency problems.
It was attempted[2] to strike a balance between reporting these broken dependencies so that they can be fixed and guarding the list of packages on a user's system as private information.
A divergent sub-thread delved[3] into the appropriate use of Conflicts: in rpm packages.
[1] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg01161.html
[2] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg01171.html
[3] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg01349.html
Making `updates-testing' More Useful
The means to enable PackageKit to prompt willing users to install testing updates was explored in a thread opened[1] by Matthias Clasen: "Basically, PackageKit should know that these are testing updates, and should ask me 'There are ... package updates available that need testing. Do you want to test these now ?' For extra points, we could even show a 'report back' link somewhere that allows to send comments to bodhi."
Richard Hughes prototyped a solution but worried[2] that it would be necessary to make changes to the users' repository configurations without their explicit consent.
A sub-thread discussed[3] the problem of out-of-sync mirrors and the use of the --skip-broken option with yum (see also this same FWN#156"YUM: Enable --skip-broken by Default?".)
[1] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00925.html
[2] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg01063.html
[3] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg01314.html
Fedora Suckage ?
The tinder for this week's massive flamewar was laid[1] by Robert Scheck in the form of a dryly ironic, multiple-topic rant. Robert attacked the use of "memory wasting" python daemons, lags in pushing updates compared to the EPEL repositories, lack of information on the recent intrusion, poor German translation, the minimal requirements for LiveCD usage, RPM-4.6 bugs, Red Hat employees blocking Merge Reviews, PackageKit bugs, and the EU support organisation for Fedora[2]!
Although there were several worthy attempts to make use of the above material for a true conflagration in general the opportunity was wasted and instead several rational, civil discussions of possible underlying causes and explanations took place. There were some worthy attempts to respond to all parts of this portmanteau complaint, but for the most part the discussion fractured naturally into several threads.
One such thread was concerned with the pushing of a D-Bus update which broke many applications including PackageKit. Kevin Kofler argued[3] that "[...] we need to be more careful with certain types of security updates, and better let them get some QA even if it means the fix gets delayed." Michael Schwendt asserted[4] the lack of active Quality Assurance as one of the contributing factors. KevinKofler explained[5] that the package had been rushed out "Because it was deemed a security update, complete with a CVE ID[.]" See this FWN#156 "The D-Bus Problem" for more details.
Max Spevack took up[6] the complaints about Fedora EMEA and more of that discussion continued[7] on the more appropriate @fedora-ambassadors list.
No further information on the security intrusion was forthcoming from Paul Frields but he relayed[8] that the matter was not being forgotten or hushed up and that he planned to meet with others to discuss communication procedures for any possible future intrusions.
Richard Hughes asked[9] for specific bugs to be filed instead of general rants: "[...] I think you need to write much shorter, to the point emails. Ranting doesn't have much affect on anything, whilst filing bugs and getting involved upstream does." He also corrected Robert that many of the daemons which he complained about were written in C, not in Python.
Colin Walters issued[10] a mea culpa: "Just to be clear, the direct push into stable is my fault; not Red Hat's or other DBus developers or anyone else's. I had originally listed it for updates-testing, but then changed the update to security and in a moment of total stupidity also changed the listing for stable."
The idea of "repeatable updates" was raised[11] again by Les Mikesell and critiqued for want of a practical implementation by James Antill. Jesse Keating made[12] a suggestion: "Treat rawhide as your 'new code' land, leave the release trees as your 'testing and working' code. That is don't be so goddamn eager to push new packages and new upstream releases to every freaking branch in existence."
Behdad Esfahbod tackled[13] the issue of Red Hat employees allegedly stalling on merge reviews. Behdad criticized the jumbling together of so many issues and repudiated any suggestion that as the maintainer of un-reviewed packages he "[...] must incorporate the merge reviews and close them, no thank you, I don't mind not maintaining anything in Fedora, and I certainly didn't block anyone from making progress in the merge reviews. When you say The Red Hat people have to follow the Fedora packaging guidelines and rules same as the Fedora folks', does it mean that Fedora should feel free to decide what *I* work on, when it doesn't decide what other Fedora folks' work on? That doesn't feel right."
The criticism of LiveCD localization was handled[14] by Jeroen van Meeuwen and he accepted that it would be useful if there were some manner in which the Spin SIG could create spins and torrent seeds outside of Fedora release engineering. It seemed that the need to make absolutely certain that such torrents and spins are kept available for support purposes may make this difficult.
[1] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00722.html
[2] EMEA is a non-profit organization with the mission to provide a focal-point and economic base for the European Fedora community. http://fedoraproject.org/wiki/Ambassadors/EMEA
[3] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00733.html
[4] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00753.html
[5] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00855.html
[6] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00772.html
[7] https://www.redhat.com/archives/fedora-ambassadors-list/2008- December/msg00092.html
[8] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00773.html
[9] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00798.html
[10] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00812.html
[11] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00832.html
[12] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00913.html
[13] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00834.html
[14] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00899.html
Help Needed: Sift "rawhide" for .pc Files
Jesse Keating requested[1] "[...] somebody to examine all the packages in rawhide that provide .pc [pkg-config] files and ensure proper placement of them based on the review guideline. This will likely require interaction with the packages maintainer(s) so the first step should probably be to produce a list of packages that ship .pc in a non -devel package and send the list (sorted by maintainer) to here so that we can discuss and pick off items."
Michael Schwendt helped[2] to start the process by providing some lists of non-devel packages which included .pc files or had requires which pulled in packages which provided .pc files.
[1] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00612.html
[2] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00648.html
Offtrac
An itch scratched[1] by Jesse Keating was to be able to interact with Trac via the commandline to create milestones for the Fedora 11 release cycle. He implemented his own python library, named Offtrac, to interact with trac using XML-RPC and asked for help in firming up the API and extending his client. Later Jesse explained[2] that the purpose was to "[...] make some aspects of using trac easier for folks, not just project owners but people who file tickets in track, like say for package tagging requests, or blocks, or... "
[1] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00738.html
[2] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg00808.html
Updates QA and Karma
The updates system came in for some more questioning (see this FWN#156 "Making `updates-testing' More Useful") when Orion Poplawski showed[1] that an rpcbind update for Fedora 9 may have been pushed to stable despite comments made by him indicating that it failed due to a dependency. Orion asked two questions: "[1] Should update submitters be allowed to give positive karma to their updates? Seems like that they are too biased. [2] Is there any requirement that an update have positive karma before being pushed to stable?"
It appeared that ultimately monitoring of such pushes are down to package maintainers and depend upon the good judgment of those doing the updates. Michael Schwendt provided[2] an overview of the situation.
[1] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg01298.html
[2] https://www.redhat.com/archives/fedora-devel-list/2008-December/msg01427.html