State of the Penguin

[unixronin]

Or, progress with Gentoo.... (posted from Slackware)

Much of the Gentoo install is up and running.  Sound doesn't appear to start on its own, and installing sound drivers apparently doesn't actually create audio device nodes .... guess you're on your own for that?  Seems strange.  Getting sound started appears as easy as a 'modprobe snd' though, and since it's the same hardware, I ASSUME I can just copy the device nodes from the Slackware install.

Speaking of device nodes, I've got udev largely beaten into doing more or less what I want it to, but a lot of it is still ugly, and I still miss the ability of devfs to dynamically create device nodes on the fly for just the hardware that's actually present.  I dislike having /dev filled up with device nodes for nonexistent hardware devices against the slight possibility that I might someday install one or two of them.  I firmly believe that the contents of /dev should accurately represent the actual hardware that is physically present -- which is why I installed devfsd in the first place.  (Having it give me SCSI device nodes that actually consistently pointed at a specified target on a specified bus on a specified host instead of "uh, y'know, like, the third SCSI disk, whichever that happens to be today, right?" was an unexpected, but welcome, bonus.)

MySQL is up, but with sound not yet working, no music on Gentoo yet.  Sigh.

Getting X working has been a battle.  Gentoo currently installs X.org 6.8.2, which dies on startup with a multiply defined symbol unless you accept the default USE flags which disable bitmap fonts (and the associated speed penalty of using outline fonts for everything you do).  What's more, it refuses to run my graphics subsystem (Matrox G450 driving a Sony GDM-W900) at any correct resolution; it loves the 4:3 aspect ratio, and absolutely will not run any 8:5 video modes.  Trust me, a 4:3 video mode on an 8:5 monitor is at best described as "sub-optimal".

And oh my god ... can someone please take xorgconfig out the back and quietly shoot it while no-one's looking?  AccelX's Xsetup was better ten years ago than xf86setup was, and in the intervening years, Xsetup has gotten better, while xorgconfig is worse than xf86setup was.  I am utterly gobsmacked at how excruciatingly awful it is at this stage of supposed maturity.  It deserves to be buried in an unmarked grave somewhere out in the back forty with a stake through its heart.  I would be prepared to bet that xorgconfig, all by itself, represents a significant barrier to wider adoption of Linux.  I mean, come on, folks, it's not just bad, it's AWFUL.

In despair at the inadequacy of X.org after wrestling with the damn thing for several hours each in two different sessions, I downloaded a demo of Xi Graphics' AccelX Summit DX Platinum.  Oh, joy, it Just Works.  It was up and running in five minutes, and five minutes after that I had it running all my 8:5 video modes, in 24-bit color with 8-bit overlays, flicker-free refresh in all modes, fast, bright, sharp, just like I've always come to expect from AccelX.  The folks at X.org should be made to buy a copy of AccelX and install it, just to see how much better xorgconfig could be.

To be fair, I've got two minor bugs with the AccelX Summit config right now. One is that some fonts aren't being rendered well.  I think starting up xfs (or xfstt, if there's still a distinction) will fix that.  The other is that cursors aren't always being changed when they should, sometimes leaving me with an incorrect cursor and guessing at where the hotspot is.  I have a support query into Xi Graphics on that.

Printing isn't up yet, because the openoffice2 ebuild mandates use of CUPS.  (This is questionable for several reasons.)  CUPS won't talk to my HP Laserjet 4M Plus; every attempt to print so far has yielded either nothing, or "WS INVALID PERS" from the printer.  It also appears CUPS is willing to speak only PCL to it.  (After all, all HP printers speak PCL, right?  ...What do you mean, you bought a PostScript Level 2 module for it?  What on earth would you want PostScript in a HP printer for?  HP printers speak PCL.  If it's good enough for us, it should be good enough for you.)

ICBM Just Works.  No problems at all.  sirc is a bit more of a problem ... sirc itself works fine, but its front-end, ssfe, is not happy on Gentoo.  If, running sirc+ssfe on the Slackware install that's running now, you type veni, vici, then use the arrow keys to back up and insert the vidi, then hit enter, you will send:

veni, vidi, vici

Do the same thing on Gentoo and you'll send:

veni, vidi@@@@@@

I haven't isolated the problem yet.  I suspect it's some kind of termcap/curses/ncurses bug, possibly in ssfe.

procmeter3 also has problems.  It's not so much a gentoo issue as a kernel issue, I think; the problem appears to be that some information in /proc/stat and /proc/meminfo has been moved and/or reordered between kernel 2.4 and kernel 2.6, and procmeter3's parsing code is written in such a way that it doesn't currectly parse lines if they're in a different order than it expects.  (Naturally, it doesn't find them at all if they're in different nodes than it expects.)  I suspect I'll be porting procmeter3 to kernel 2.6.  Whee, a project!

I had to change all the accelerators in Much To Do About Nothing, my Perl/Tk to-do list program, from Alt+key to Ctrl+key because Alt+key appears not to work on [either Gentoo or AccelX Summit].  I don't yet know why this is.

Mail isn't up yet on Gentoo.  I'll probably be leaving Postfix and Dspam to be among the later things I set up.  Apache isn't up yet either.  I suppose since Gentoo, with the packages I have installed thus far, seems bound and determines that it wants to use Apache2, it's time to figure out how to port my Apache configuration to Apache2.  At least with Apache2, as I understand it, I no longer need to separately install mod_ssl, because it's effectively built in.  (Is this also true of mod_perl?  If not, can anyone with Apache2 and Gentoo fu tell me what I need to do on Gentoo to get mod_perl into Apache2?)

I installed the Gentoo LogJam package.  It's 4.4.1, and appeared completely broken.  So, with prerequisites in place, I uninstalled it and built my own enhanced version of 4.5.1 from source by hand.  With libgtkhtml finally present (I gave up wading through the appalling Gnome dependency tree on my Slackware installation), I can finally use Logjam's preview.

....Er ..... wait one fucking minute ..... you call that a preview?  That's not a preview!  None of the CSS is rendered!  I wasn't missing much, was I?

Comments

[tww1fa]

Make sure that RC_DEVICE_TARBALL="no" in /etc/conf.d/rc - otherwise Gentoo will pre-populate /dev with a default set of devices (fail-safe for people who don't have udev set up properly.) Turning RC_DEVICE_TARBALL off makes sure that only existing devices are listed.

For sound, "alsaconf" should set up the module stuff, then you do "rc-update add alsasound default" and run "alsamixer" to unmute.

I have absolutely no problems talking to my Laserjet 4M Plus from CUPS - I have it set up both for native Postscript as well as a Foomatic->ghostscript->PCL chain (this chain tends to be much faster than the printer's built-in postscript for bitmap-heavy printing.) Printing is through JetDirect using socket://lj4mplus in all cases. I use the PPD from http://www.linuxprinting.org/foomatic-db/db/source/PPD/HP/mono_laser/HP_LaserJet_4_Plus.ppd - just drop this in /usr/share/cups/model and you should then see "HP LaserJet 4 Plus v2013.111 Postscript (recommended)" added to the list of HP printers. The other option is to set the "ppds" and "foomaticdb" USE flags so that PPDs will be installed for more printers than you can shake a stick at (probably overkill in your case.)

[darthgeek.livejournal.com]

As far as CUPS and Laserjet4's, there's a log that CUPS writes to (I've forgotten where), but it's very much in the veign of apache's access and error logs. When I tried to print to the lj4 at work, it wrote a very clear error to the error log telling me that I needed foomatic (much like tww1fa said)

[unixronin.livejournal.com]

Yeah, I think I have the CUPS issues sorted. I'm about to reboot into Gentoo and try it again, soon as I can get xfs to install.

[attutle.livejournal.com]

"I would be prepared to bet that xorgconfig, all by itself, represents a significant barrier to wider adoption of Linux. I mean, come on, folks, it's not just bad, it's AWFUL."

Sounds like you ought to try Fedora or CentOS, so you have a better sense of which problems relate to Linux and which ones relate to particular distributions. It might have details you don't like, but it is quite trivial to get running. The problems you are having sound more like barriers to wider adoption of Gentoo.

[unixronin.livejournal.com]

xorgconfig isn't going to come in different flavors depending on the Linux distro X.org is being shipped with. If CentOS, say, has a better config tool that bypasses or is a front end to xorgconfig, that doesn't absolve xorgconfig from being a hostile piece of junk. Xi Graphics can create a decent, easy-to-use, powerful setup tool and ship it with their Xserver; why can't X.org?

[attutle.livejournal.com]

But that's just the thing. Why are you interacting with xorgconfig in the first place? What did Gentoo fail to do such that you ever even because aware of it?

Why didn't X just work, like it does with Fedora/RHEL/CentOS/Knoppix/...?

Did I just see someone post...

[msa.livejournal.com]

"Why didn't $FOO just work, like it does with Fedora/RHEL/..."

In my experience, very few values of $FOO 'just work' with a RedHat product.

Re: Did I just see someone post...

[attutle.livejournal.com]

I was going to see what I could learn about you from the page listed in your profile, but it doesn't work for me. Is that because I'm using Fedora? ;-}

I'm wondering if our experiences are really all that different, or if you're just bashing old versions.

My experience is that most things work pretty well for common values of $FOO, which seems relevant in a discussion about "significant barriers to wider adoption".

They work well enough that xorgconfig is pretty far off the beaten path.

Re: Did I just see someone post...

[unixronin.livejournal.com]

I guess my viewpoint is (a) the fact that some $OS distros provide a better configuration method that saves people from having to use xorgconfig doesn't excuse X.org from putting out one that's so awful, and if X.org put out a decent config tool, ALL distros would be able to have five-minute X/11 setup without needign to develop their own tools to do it.

Re: Did I just see someone post...

[attutle.livejournal.com]

To be fair, there was a time when it was difficult to get patches accepted into the XFree86 tree. ;-}

But I still don't understand why you were using xorgconfig in the first place.

Gentoo's X Server Configuration HOWTO (http://www.gentoo.org/doc/en/xorg-config.xml) suggests using Xorg -configure, and only lists xorgconfig as an alternate.

Re: Did I just see someone post...

[unixronin.livejournal.com]

Xorg -configure was barfing with multiply defined symbols.

[smandal.livejournal.com]

(Having it give me SCSI device nodes that actually consistently pointed at a specified target on a specified bus on a specified host instead of "uh, y'know, like, the third SCSI disk, whichever that happens to be today, right?" was an unexpected, but welcome, bonus.)

udev on Gentoo now does this by default. Look in "/dev/disk/by-path". I'm using udev-073.