Now I know why CentOS can’t upgrade its XFCE from Extras:

This is the binary installer for XFCE 4.6.0, because there isn’t any for 4.6.1 as of yet, and it can’t install because in EL5 gtk-doc is 1.7 < 1.10. Holy shit. XFCE 4.6.x can’t run on EL5. What a pile of shit… EL5!

Another annoyance is that my X session crashed one more time in relation to multimedia. Some time ago it crashed when I inserted a DVD and I tried to used VLC to play it (because no matter what I tried, Totem can’t play DVDs in EL5); upon restarting X, the DVD was able to play smoothly, but for the price of crashing X. Today, my X crashed when I right-clicked on an AVI file and I told it to play with one of the media players, I guess it was Kaffeine, but this is not relevant; after restarting X and Kaffeine, opening the file made it play flawlessly.

I wish to the developers of the Intel video drivers for Linux to burn in Hell, after having been torn in many pieces by hungry vultures! There is nothing wrong in the hardware, because it doesn’t crash Windows; but it crashes X on occasions.

A last annoyance with EL5 is the need for me to periodically «killall npviewer.bin». That’s because I normally want to be able to view all the Flash content I might encounter, hence I use the Adobe Flash plugin into the moronic Firefox. That makes two more piles of shit: one from Adobe, and the other one from Mozilla.

I should do something to preserve what was left out of my nervous system. Expect Odiecolon.repo to be frozen soon, I’m losing interest in hacking EL5 clones…

Have you ever encountered the situation where you open an ISO image with FileRoller (or you mount it with Nautilus), only to see a fake readme file saying this?

This disc contains a “UDF” file system and requires an operating system that supports the ISO-13346 “UDF” file system specification.

Fsck. Not only I encountered this in CentOS 5.3, but it seems that even the latest GNOME from the all-mighty Ubuntu is dumb, so you have to mount the ISO file manually, or to use handcrafted scripts. Unless you mount the file manually, the cretinoids who developed the marvelous open-source stuff thought you must be wanting to mount ISO9660 and nothing else! Duh…

Now, say I mounted the ISO by hand. How the fsck can I edit it in Linux, e.g. to remove, rename or edit a few files? It’s freaking bootable and it must keep being so, I can’t just copy the files elsewhere!

Not even the “powerful” K3b can help: with ISO files, it can only burn, not edit them. And no, the “magical” IsoMaster is of no use: once I open the bloody ISO file…

This disc contains a “UDF” file system and requires an operating system that supports the ISO-13346 “UDF” file system specification.

Fsck. Useless. (Try it yourself.)

Oh, what was that “special” ISO file? A DVD image. Say… umm… err… uh… something like Windows 7, yup. What, can’t I try it? I just wanted to rename the file ei.cfg before burning it, so I could get a menu allowing me to select which version to install (specifically, I wanted Home Premium).

But Linux is too dumb for that.

UPDATE: Oh, it looks like I could use mkisofs to regenerate a modified ISO! Would it work with a Windows 7 boot image too?! Would the DVD work as expected?

REUPDATE: As per here, I tried to recreate an ISO from the mounted directory something like this:

mkisofs -b "boot/etfsboot.com" -no-emul-boot -boot-load-size 8 \
-iso-level 4 -allow-lowercase -J -l -D -N -joliet-long -relaxed-filenames \
-V "GRMCULFREO_EN_DVD" -o ../win7.iso .

…but the resulting image was not bootable (”Couldn’t find BOOTMGR”)! Strange thing, the reconstructed image was “not UDF enough”, because ISO Master was able to fully load it!

If you weren’t already aware of that, Red Hat sucks big (I was tempted to write that they suck my dick, but maybe my dick is not big enough to match Red Hat’s suction capacity). One more time, they find tremendously difficult to fix a bug that was reported back in January and that cripples some basic existing functionality of EL5.3!

There is a post on the CentOS ML that reminded me that puplet doesn’t tell me anymore that I have updates in 5.3. That’s the upstream Red Hat Bug 480339 - puplet does not show updates, reported on 2009-01-16 and still “NEW” (it might get fixed in EL 5.4 though), albeit marked as…  Priority: urgent, Severity: high.

This is a known issue with 5.3, and it has been listed in the CentOS 5.3 Release Notes, however there was no real workaround so far — except for using yum to check for updates.

Johnny Hughes has (surprisingly!) decided to be more active than expected, and he has provided an interim patch, which is an upgraded dbus-python-0.82.

Red Hat, you suck.

dbus-python-0.82.4 was also added to Odiecolon.repo (32-bit and source).

Suppose you’re from Egypt. You’re now back to winter time, but does your OS know about that? If not, and if you “fix” the wrong hour the wrong way, it’s like you’re on the wrong timezone, and this will affect your communication (e.g. e-mail). Heck, we’re in a globalized world!

Red Hat Enterprise Linux has provided the updated tzdata-2009k on August 17. Scientific Linux provided it the next day and, surprisingly, CentOS was this time even faster!

How about other major names? How about Debian, known for only pushing tzdata updates in debian-volatile? Well… we’re now 3 days after the timezone change in Egypt and the corresponding update, tzdata_2009l, is not in debian-volatile, but only in Debian testing and unstable! So no, Debian Lenny and Etch don’t have it, not even as a “volatile” update.

Slackware missed it too, in both 12.2 and current. Too bad.

Oh, did I mention how Ubuntu Hardy, Intrepid, Jaunty and Karmic got the Egypt timezone change with tzdata_2009i, and they now have even newer timezone files, to incorporate some more fixes?
» Read more…

Emesene 1.5 does build (so to speak; it’s Python), but it can’t run on EL5 (most likely it needs a newer Python):

File "/usr/lib/emesene/ConversationUI.py", line 596
  template[0] = ''.join(x if i % 2 == 0 else '' for i,x in enumerate(template[0].split("$")))
                           ^
SyntaxError: invalid syntax

Too bad, as the announcement said: «No more gui locks».
Users of Mandriva Cooker can find it in contrib/backports.

That’s an easy one:

• Added fontforge-20090622. Still ugly as a software though.
• Added gnome-specimen-0.3.

From Null pointers, one month later:

Red Hat has not, as of this writing, issued an update for this vulnerability. That is unfortunate because most RHEL systems are vulnerable as the result of a policy choice made by Red Hat. RHEL systems, by default, allow “unconfined” users to map low addresses addresses. Red Hat’s Dan Walsh explains: “We are not planning on changing the default in RHEL5, to maintain backwards compatibility.” So, because compatibility trumps security, RHEL systems (and those running distributions based on RHEL) remain vulnerable to a trivial local root problem with exploit code easily available and in use. Not good.

I didn’t bother to check this thoroughly, but is this the same kernel vulnerability that was described a month ago as «A vulnerability which, when viewed at the source level, is unexploitable! But which, thanks to gcc optimizations, becomes exploitable :)», or is it a new one? Anyway…

• New Linux Kernel Flaw Allows Null Pointer Exploits
• Another kernel NULL pointer vulnerability
• Linux NULL pointer dereference due to incorrect proto_ops initializations (CVE-2009-2692)
• Eight Years of Linux Kernel Vulnerable
• Linux Kernel ’sock_sendpage()’ NULL Pointer Dereference Vulnerability
• Red Hat Bug 516949 -  (CVE-2009-2692) CVE-2009-2692 kernel: uninit op in SOCKOPS_WRAP() leads to privesc

What’s another term for «false sense of security»? SELinux?

Possible temporary workarounds found on the main SL mailing list: here and here.

In brief:

• Updated Sylpheed to 2.7.1.
• Added google-translate (ASTranslator, a Qt4 applet: screenshot)
• Added ncftp-3.2.2
• Added duplicity-0.5.18
• Added ssmtp-2.6.3 (to relay through a GMail account: basic settings in /etc/ssmtp/ssmtp.conf, then test with: /usr/sbin/ssmtp recipient@domain.com < bodyfile)
• Added xdialog-2.3.1 (built for Gtk+2 only; basic tutorial, looks better in Gtk+2)

As usual, yum clean all is highly recommended for a guaranteed update of the repodata.

Hopefully there aren’t issues to be fixed, as this Friday I’m out of town.

In brief:

• Updated gxine to 0.5.904 (don’t ask me what’s new, I don’t know).
• Added p7zip-9.04 + p7zip-light-9.04.
• Added gdm-themes (extra GDM themes, useful especially for Scientific Linux).
• Minor fixes leading to rebuilds of xine (build options) and acidrip (cosmetic fix in the spec file).

As usual, yum clean all is highly recommended for a guaranteed update of the repodata.

NOTE: Older versions of the packages (e.g. gxine-0.5.903) are not stored by the repository for space reasons (currently, the 719 RPMs take about 0.9 GB), thus making sure the repodata is updated is essential. When in trouble… and in doubt… yum clean all.

Although nobody bothers to tell me whether the repo works for them or not, I’ve decided to add a few more applications to it:

• geany-0.17 and a number of geany-plugins-*-0.17.1
• bluefish-unstable-1.1.6 (because 1.3.x requires a newer GTK+, and 1.0.7 is just too old; it can be used to browse via SFTP/SMB/FTP, read this old post)
• pida-0.5.1 (using vim-X11)
• gazpacho-0.7.2 (en passant)
• sbackup-0.10.5 (not for advanced users)
• webmin-1.485 (it allows you to completely screw your system)

As usual, yum clean all is recommended for a guaranteed update of the repodata.

I might be out of town for today…

Ms. Martin, we know her. Surprisingly for someone who makes a living partly from the RHEL expertise, she initially decided that CentOS is not good for the mere fact that a stupid netbooks (aren’t all the netbooks stupid? 1024×600 in the 21st century?) is not fully supported for a clone of Enterprise Linux. Cool, eh?

I am going to pardon the fact that, for some reason, she decided to act strangely mad. Was this because two guys dared to dislike Vector Linux? She told to the 2nd guy: «You wrote: I am an idiot. This part I agree with.»

She actually provided me with an information I missed: she counted 70,000+ packages in Ubuntu. The last time I was interested in how many packages there are in Ubuntu, they were more like 20,000-22,000. Today, things have changed: not counting the source packages and the d-i ones, I’ve got something like 34,000+ packages in Jaunty. So maybe they’re 70,000 deb files when you count sources & everything. Still, Ms. Martin was not fair with this one.

Never mind. The entropy continued to increase, because life played Ms. Martin’s game. The End of the CentOS Netbook Experiment is radically negative, et pour cause:

» Read more…

Added to Odiecolon.repo:

• PyChess 0.8.4
• PyTraffic 2.5.4 (screenshot)
• belooted-0.1.4 (jeu de belote: EN, FR; screenshot)
• Gnome Hearts 0.3
• ghextris-0.9.0 (screenshot)
• pdfedit-0.4.3

As usual, yum clean all is recommended for a guaranteed update of the repodata.

Human-readable changelog:

• Added Sylpheed 2.7.0, a great lightweight mail client with fewer dependencies (and fewer plugins than Claws). BTW, it knows to thread the summaries, something Thunderbird only learns with version 3. Added as a dependency: GnuPG 2.
• Added Mercurial 1.3.1.
• Added Gnaural 1.0, an old faiblesse of me.
• Added GnoCHM 0.9.11, despite being much slower than xCHM (also in this repo).
• Added AcidRip 0.14, which is simple and practical (and has fewer Perl dependencies than other popular DVD rippers). Note that I had to use lsdvd-0.15 instead of 0.16, so I rebranded it as lsdvd-1.15, because otherwise AcidRip can’t process the list of the DVD contents on EL5. It happened to other people too.
• Added gnormalize-0.6. Added as a dependency, mppenc/mppdec (Musepack support).
• Added dar-2.39 + KDar 2.10 (backup).
• Added Luma 2.4 (LDAP GUI tool).
• Added Trickle 1.07 (bandwidth limiter).
• Added fpc-2.2.2 + Lazarus 0.9.26.
• Added Sweep 0.9.3 (homepage).

yum clean all is recommended for a guaranteed update.

Right now, the repository has the following number of RPMs: 235 in SRPMS, 35 in noarch, and 396 in i386. The grand total is… 666. Wow. I feel diabolique…

Yesterday’s scandal simply left me gaping (bouche bée, actually). This can’t be… but it is. Lance Davis, one of the founders of the CentOS Linux project and owner of the domain centos.org, also director at UK Linux Ltd. and (presumably still) owner of Cheeplinux, is… Missing In Action… Absent Without Official Leave… unreachable, although suposedly not dead, nor seriously ill…

…and he’s holding the entirety of the donations that ever went to the CentOS project, if my understanding is correct. Gulp.

There is an Open Letter to Lance Davis posted on the front page of their site, on the mailing list, and on the developers’ personal blogs: one, two, three… generally you should find them on Planet CentOS.

The issue even got on Slashdot, and somebody updated the Wikipedia page on CentOS!

Dag added an interesting post on The burden of keeping things private, which is however weird, given that he personally resigned from the CentOS team one month ago.

Other reactions include: a thread on LinuxQuestions.org, a summary page on Mahalo, and probably a lot of opinions on personal blogs will follow.

The initial discussion on the mailing list has over 50 messages as I am writing this, not counting a split thread and a panic thread.

Very much unlike me, I have no personal stance on the matter. I have incidentally exchanged a couple of e-mails with Lance Davis a couple of years ago, but that was all: I don’t have an opinion on the guy.

I however suspect that, once Scientific Linux 5.4 released (hopefully with a nicer GDM theme), many, many new installations are going to be SL instead of CentOS. They have performed well lately, they’re the first to release a clone of EL 4.8, and also the first EL clone to rebuild Firefox 3.0.12.

Frankly, the way CentOS lags is not Lance’s fault.

UPDATE, AUG. 1: It looks like Lance Davis is going off the stage amiably…

Just a couple of them:

• Added the missing dependencies (otherwise, EPEL has them too) libsigc++20, glibmm24, gtkmm24, cairomm to fix GParted and Eiciel.
• Added soundconverter-sl for the users of Scientific Linux, who can’t install the regular soundconverter. It will pull gstreamer-plugins-extra (which has MP3 and tag support) instead of gstreamer-plugins-ugly (needed in EL and CentOS).

I definitely would like more success or failure reports from other users of EL5 clones…

UPDATE:

• Added gstreamer-plugins-ugly-sl-0.10.9 - THIS IS FOR SCIENTIFIC LINUX ONLY! SL uses a modified gstreamer-plugins-extras which conflicts with gstreamer-plugins-ugly (it should conflict with RPMforge’s and RPM Fusion’s too!) because they both include libgstlame. SL users need to install this package instead of the regular gstreamer-plugins-ugly.

Sort of (dependencies not always mentioned):

• Added libmms-0.4 to fix gnome-plugins-bad, and rebuilt the latter.
• Added gnome-commander-1.2.7 (newer versions won’t build).
• Added Dillo 2.1.1 (and FLTK2 en passant).
• Added aria2c 1.3.1.
• Added emesene-1.0.1.
• Added emelFM2 0.6.0.
• Added diffuse-0.3.4 (as an alternative to Meld).
• Added dosbox-0.73 (works great with the MS-DOS version of Pushover).
• Added GParted 0.3.9.
• Added PyRenamer 0.6.0.
• Added SoundConverter 1.3.2.
• Added AlsaPlayer 0.99.80 (+alsaplayer-fft, alsaplayer-mac).
• Added mac-3.99 (Monkey Audio, CLI).

KNOWN ISSUE WITH SCIENTIFIC LINUX 5.3: Because SL uses a modified gstreamer-plugins-extras (to make use of lame, taglib, libid3tag, libmad), it conflicts with gstreamer-plugins-ugly (it should conflict with RPMforge’s and RPM Fusion’s too!) because they both include libgstlame. This makes SoundConverter impossible to install, as it requires gstreamer-plugins-ugly.

I am afraid coffee can not replace a good sleep (that’s why everybody wants to be a cat), but I have to deal with what I have…

• There is a guy telling you: Don’t buy Linux laptops (atl1c fail), «…and don’t buy Acer laptops!». Yeah, sure. Don’t buy Dell laptops. Don’t buy HP laptops. Don’t buy Toshiba laptops. Don’t buy ‘[[:alnum:]]+’ laptops. Because each and every vendor has at least one model that sucks. The key is to know what you’re buying, not to boycott an entire brand… otherwise you should boycott them all! My cheapo Acer TravelMate has a Broadcom for the wired Ethernet, and it works tremendously well. Should I win the lottery, I would rather buy a ThinkPad Lenovo though. I guess.
• Google is recursively recursive. Seen here.
• No, I cannot live without Adobe Flash, no matter how crappy it is. And no, Acrobat Reader is not bloatware (although the size of the download is huge): it’s the fastest PDF renderer (Xpdf excluded), and significantly faster than the bloody Evince. Open source is not necessarily better.
• Where the fuck has this guy found a package named QMMS in Ubuntu? The only link given is… apt:qmmp. But there is no qmms in Ubuntu, and Google can’t find anything either.
• Finally, I should recommend people Scientific Linux 5.3 for new installs instead of CentOS 5.3. The bloody CentOS team (normally, Karanbir Singh is to blame in this case) could not find the time to build Firefox 3.0.12 (a security update) from firefox-3.0.12-1.el5_3.src.rpm, posted by TUV on July 20, at 03:55. Compare this to Scientific Linux, who offers Firefox 3.0.12 since July 23, at 16:30. I might understand that KS is busy with CentOS 4.8, but FF is a single bloody package, and there are only 2 architectures, and we’re 8 days later, dammit! There is nothing fancy to test: if it’s (say) unstable, the responsibility is upstream, at Red Hat! The only problem I have with SL is their atrocious artwork, especially when comes to GDM…

Yes, she does. I couldn’t possibly assert that “CentOS (or Scientific Linux) is marvelous for the desktop” — and I have never stated such a thing. On the contrary, what I said is that RHEL5 is sort of another XP, which can be read in several keys.

It only happens that I am using it today (CentOS 5.3 on the Acer and Scientific Linux 5.3 with IceWM on the HP Omnibook). Not because CentOS is “really good”, not because it’s “better than other distros”, but merely because it’s “standard” (and Mrs Martin knows that very well) and because… everything else is worse in a way or another!
» Read more…

Humanized changelog:

• Fixed the missing dependencies for Deluge and aMule.
• Removed Git and QGit because they require way too many Perl packages. People who need more than CVS should pull whatever they need from EPEL.
• Added Pushover.
• Rebuilt GIMP 2.3.15 and 2.3.14 to appear in RepoView as Graphics, not Multimedia.
• Rebuilt GXine to require xine-lib-extras-freeworld, so you could actually hear something.
• Added AntiWord 0.37.

Hopefully, all the files are uploaded correctly. I also hope I’ve not cleaned up too many of the old files. Oh, and yum clean all shouldn’t hurt.