Fedora Weekly News #118 - Development Beat
== Developments ==
In this section, we cover the problems/solutions, people/personalities, and
ups/downs of the endless discussions on Fedora Developments.
http://www.redhat.com/mailman/listinfo/fedora-devel-list
Contributing Writer: Oisin Feeley
=== Net-install ISO Already Exists As ”rescuecd.iso” ===
MichaelDeHaan reported[1] on a discussion at FUDCon on the use of
Debian-style net-install ISOs. He posited large bandwidth savings as
one of the attractions. ChrisAdams was less sure[2], commenting that a
DVD download was used multiple times, while each net install would
require duplication. LesMikesell recommended[3] a proxy to cache large
files to those who do much of this type of thing.
[1] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03014.html
[2] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03016.html
[3] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03022.html
JasonTibbitts thought[4] that he had been doing what Michael described
for all his Fedora releases. JeremyKatz added[5] that the ”rescuecd”
image was already capable of this and was going to be renamed to
emphasize its utility. He added that ChrisLumens had been doing some
work to delay repository selection to stage2 in order to allow the
selection of mirrorlists and proxies.
[4] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03015.html
[5] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03018.html
=== Firefox 3 Liberation Fonts ===
A simple request[1] from MarkG85 to make Firefox 3 use the Liberation
fonts as defaults in Fedora 9 prompted and interesting discussion.
MarkG85 posted screenshots of his experience in Rawhide and claimed
that the Liberation fonts were obviously better than the defaults.
NicolasMailhot was highly skeptical[2] and suggested that “looking
better” was a purely subjective judgment mostly influenced by
familiarity. He also commented that encoding coverage and upstream
responsiveness were more appropriate metrics and that DejaVu won on
that basis.
[1] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg02847.html
[2] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg02857.html
FelixMiata argued[3] that if the goal of choosing fonts which matched
Windows’ font metrics were accepted then LiberationSans proved the
best substitute for Arial. The advantage of this would be uniformity
of web browsing experience. Nicolas’ detailed and interesting response
pointed[4] out that Windows fonts have varied widely across time and
application and that due to the Mozilla Foundation’s lack of
stewardship it would be best not to chase fonts. Instead effort might
be better spent on “creat[ing] our own solid font set than continue
chasing the Microsoft tail indefinitely - it has the financial means
to move way faster than us on the font creation front anyway.” This
exchange between Felix and Nicolas is worth reading. Nicolas
pointed[5] out that the page-zoom[6] features of Firefox 3 will result
in huge differences between Fedora and Windows systems even if the
very same fonts are installed and that “legal issues” cannot be
ignored.
[3] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg02862.html
[4] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg02870.html
[5] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg02953.html
[6] http://mozillalinks.org/wp/2007/07/try-firefox-3-full-page-zoom-with-full-page-zoom/
Several people argued that the premise of the thread was mistaken.
Among them was BehdadEsfahbod who outlined[7] the normal course of
events in which a web-browser picks which fonts to use. MarkG85
restated[8] his objectives to make it clear that he was not interested
in “mimic[ing] a well known other operating system” but rather trying
to recapture the pleasant appearance of Firefox 2.
[7] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg02956.html
[8] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg02871.html
A suggestion was made[9] by NicolasMailhot that the way forward
involved ripping out Firefox’s font defaults and making it use desktop
font settings so that users with a preference can implement their
choice across all applications.
[9] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg02887.html
=== Firefox And Epiphany Crashes ===
A report[1] by HorstvonBrand that Firefox and Epiphany were crashing
after the updates of 30 Jan 2008 led to a request from
ChristopherAillon that people testing Firefox 3 would report bugs
upstream instead of at Red Hat’s bugzilla. JoachimFrieben had
experienced[2] the same issues as Horst and pinpointed the changes to
a specific ”xulrunner” build.
[1] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03118.html
[2] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03135.html
Christopher explained[3] that he was pulling from CVS HEAD and in
general not adding patches to the Fedora package and it would thus be
a good idea for Fedora testers to report bugs upstream. JakubRusinek
(livio) preferred[4] to pull the nightly snapshot from Mozilla itself
although Christopher argued that there would be no final
Mozilla-distributed binary so it would be helpful to test the Fedora
binary. Jakub noted that he needed fresh vanilla sources for his
purposes, to which Christopher responded[5] that he was pulling from
CVS, packaging them in RPM format and that the nightly builds at
Mozilla were often older than Christopher’s own. Jakub still seemed[6]
to prefer using Mozilla’s builds arguing that it was easier to find
changes.
[3] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03155.html
[4] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03168.html
[5] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03187.html
[6] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03189.html
A similar point was made[7] by Horst, who asked how it would be
possible to know what Fedora-specific tweaks had been made, what the
CVS revision was and also pointed out that these would not be
“official” upstream binaries, all of which would make it difficult to
report bugs intelligently. Christopher responded[8] that: 1)no
upstream binaries exist; 2)there is no such thing as a CVS-wide
revision, so snapshot date is acceptable for bug filing;
3)Fedora-specific tweaks are minimal and involve using shared
libraries instead of static libraries in the binaries. Also of
interest is the building of Firefox as a XULrunner application about
which Mozilla is keen to receive feedback. This seemed good enough for
Horst who duly opened a bug report[9].
[7] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03173.html
[8] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03178.html
[9] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03192.html
=== Cannot Access Hardware Clock Via Any Known Method ===
Repeatedly during booting of a rawhide machine SteveGrubb noticed the
message “Cannot access the Hardware Clock via any known method” and
wondered[1] if anyone else running x86_64 rawhide had noticed the
message. RexDieter confirmed[2] that he and TomCallaway had noticed it
and speculated that ”hwclock” was being fired up before ”/dev/rtc”
had been created by ”udev”.
[1] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03042.html
[2] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03043.html
Further discussion led HansdeGoede to point out[3] that there was a
bugzilla entry which described[4] the problem and possible solutions.
It seemed that the problem is due to there being three possible RTC
drivers and ”mkinitrd” creating ”/dev/rtc” with the wrong
major/minor device numbering.
[3] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03084.html
[4] https://bugzilla.redhat.com/show_bug.cgi?id=290731
A detailed explanatory post[5] from Hans explained his view of the
problem, his proposed fix and expressed his desire that some attention
would be drawn to the bug. JoachimFrieben suggested[6] a temporary
means (synchronizing the system clock before startup with NTP) of
mitigating the bug until it is fixed. BillNottingham agreed[7] with
Hans that it might be possible to move the setting of the clock later,
but still make it available as early as possible in order to
co-ordinate log messages and other events. Using the kernel to
automatically set the clock upon ”/dev/rtc” instantiation was noted
as being limited by the requirement of all users to use UTC.
SimoSorce asked and DanHorák confirmed[8] that this was a problem for
dual-boot systems. Dan provided an interesting link to a plea[9] to
Microsoft to set the RTC to use UTC.
[5] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03047.html
[6] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03052.html
[7] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03072.html
[8] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03080.html
[9] http://www.cl.cam.ac.uk/~mgk25/mswish/ut-rtc.html
=== Devilspie And Angelcake: Asus Eee PC Continued ===
The discussion of the Asus EeePC (FWN#117 “Fedora 9 For Asus
EeePC”[1]) continued with some interesting developments. It seemed[2]
that ChrisSnook and JayCliburn, as the main Atheros driver developers,
had received a “huge code dump from Atheros” and were producing a
unified ”atlx” Atheros driver. Chris requested ”atl2” testers to
contact him off-list.
[1] http://fedoraproject.org/wiki/FWN/Issue117#head-751f643f1077b796d203768c770db644c81ea364
[2] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg02884.html
ChuckEbbert responded[3] to Orion’s earlier question with the
information that ACPI support for the Eeeepc is being worked on by
Debian developers and discussions on upstreaming it into the kernel
are happening.
[3] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03097.html
LamontPeterson took up[4] JohnPalmieri’s suggestion that avoiding a
swap partition, while desirable, might lead to more OOM[5] troubles.
Lamont recalled a Linux Journal article on the topic of setting memory
ceilings on embedded systems to prevent the OOM Killer triggering.
[4] https://www.redhat.com/archives/fedora-devel-list/2008-February/msg00043.html
[5] http://en.wikipedia.org/wiki/Out_of_memory
[6] https://www.redhat.com/archives/fedora-devel-list/2008-February/msg00078.html
A quick note posted[7] by OrionPoplawski about the dearth of
information on the Eeepc (and his own identification of the drive as
“SILICONMOTION SM223AC”) stimulated JonNettleton to add[8] that he had
been hacking on ”devilspie”[9] (a utility to detect window creation
and apply placement and sizing rules to them) in order to make it
automatically resize windows in response to window-geometry
thresholds. RahulSundaram shared[10] a current discussion with the
Eeedora developers who seem to be interested in making their distro
work on the 2GB Eee PCs.
[7] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg02928.html
[8] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg02933.html
[9] http://www.burtonini.com/blog/computers/devilspie
[10] https://www.redhat.com/archives/fedora-advisory-board/2008-January/msg00277.html
Jon clarified[11] that he was not part of Eeedora[11a], and was
primarily interested in a heavily-changed, non-Fedora-branded
derivative with many changes to upstream Fedora. His central
interests involve dealing with the challenges of the small,
low-resolution screen: hence his hacking on ”devilspie” and a
possible GUI tool (provisionally named ”angelcake”) to configure it.
Jon also expressed interest in updating ”asus_acpi” to handle the
keys and improving boot speed. YaakovNemoy suggested using XMonad[12]
instead of ”devilspie” but Jon was able to point to the ability of
Metacity to do compositing (still in Rawhide) and the simplicity of
”devilspie”. DouglasMcClendon also seemed taken with it and
suggested[13] investigating ”wmctrl”.
[11] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03061.html
[11a] http://wiki.eeeuser.com/howto:eeedora
[12] http://xmonad.org/
[13] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03180.html
After all the interest shown Jon was encouraged by RahulSundaram to
start a wiki-page detailing his progress, which seems[14] very
promising.
[14] https://www.redhat.com/archives/fedora-devel-list/2008-February/msg00078.html
=== PulseAudio Crashing Applications? ===
A request was made[1] by WarrenTogami for other people’s experiences
with ”pidgin”, ”mplayer”, ”xine” crashing on a regular basis,
especially whether they had backtraces. MatthiasClasen wondered[2] why
Warren assumed that his crashes had anything to do with PulseAudio if
he had not yet seen backtraces.
[1] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03123.html
[2] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03124.html
Some kvetching about PulseAudio in general and why it had been
included in Fedora 8 stimulated ChrisSnook to complain[3] about the
general state of KDE in Fedora. He was clear that he welcomed the
“Gnome developers […] pushing ahead with all this nest stuff[.]”
When AndrewFarris pushed[4] him on the point, suggesting that the
problems were due to the small number of Fedora KDE developers rather
than Gnome developers breaking KDE Chris produced[5] a splendid rant.
His general point seemed to be that KDE development was too hampered
by interference from “Gnomifiers” and that there was a lack of
responsiveness compared to, for example, SELinux development.
[3] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03138.html
[4] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03140.html
[5] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03142.html
The wind was slightly taken[6] out of the sails of the complaint when
RexDieter explained that the packaging of PulseAudio had been
constructed carefully so that it was easy to remove it in KDE with a
{{{rpm -e kde-settings-pulseaudio alsa-plugins-pulseaudio}}} followed
by restarting the KDE session. LaurentRineau added[7] some suggestions
about how this could be made a per-user choice.
[6] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03148.html
[7] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03162.html
Backtraces were supplied[8] by AndrewFarris of a crash of
”rhythmbox” and also some possibly related to ”xulrunner”.
[8] https://www.redhat.com/archives/fedora-devel-list/2008-January/msg03141.html
