Miskatonic University Press
William Denton <wtd@pobox.com>
I've been using FreeBSD since
1999. I like it a lot. Right now I run 6.x and am tracking
-STABLE on three machines. I upgrade them every month or
two.
ping: sendto: Permission denied
Yesterday I upgraded widmerpool to the latest 6.2-STABLE. Everything
went fine except that the machine refused to get on the network.
Absolutely no packets were getting into or out of the box. Pings gave
the error ping: sendto: Permission denied. I knew it was
something related to the firewall I set up in
March, but the problem persisted even after I set
ipfilter_enable="YES" in rc.conf.
The
FreeBSD Handbook says, "ping: sendto:
Permission denied error messages are often caused by a
misconfigured firewall. If ipfw is enabled in the kernel
but no rules have been defined, then the default policy is to deny all
traffic, even ping requests!"
That's what I had. I added a line to my kernel config:
# Firewall options IPFIREWALL options IPFIREWALL_VERBOSE options IPDIVERT # Enable NAT options IPFIREWALL_DEFAULT_TO_ACCEPT
Now by default everything is allowed through. I immediately disallow everything in my firewall config file, and then permit particular kinds of traffic.
I got fed up with the way xterm was showing non-English characters. Sometimes even simple stuff like é, à, and ø didn't display properly. I installed rxvt-unicode, which is urxvt in the ports. It works very well, and now when I look at this UTF-8 demonstration text file it (almost) all makes sense.
I put these lines in my .Xdefaults, mostly copied from
the sample on the web site. I'm not sure what they all mean, and they
don't stop my lynx sessions from having coloured text on a
black background, which is horrible, but I'll get that solved soon.
urxvt.font: -*-courier-medium-r-*-*-14-*-*-*-*-*-*-*,\
-misc-fixed-bold-r-normal--15-140-75-75-c-90-iso10646-1,\
-misc-fixed-medium-r-normal--15-140-75-75-c-90-iso10646-1, \
[codeset=JISX0208]xft:Kochi Gothic, \
xft:Bitstream Vera Sans Mono:autohint=true, \
xft:Code2000:antialias=false
urxvt.boldFont: -xos4-terminus-bold-r-normal--14-140-72-72-c-80-iso8859-15
urxvt.italicFont: xft:Bitstream Vera Sans Mono:italic:autohint=true
urxvt.boldItalicFont: xft:Bitstream Vera Sans Mono:bold:italic:autohint=true
URxvt.background: BlanchedAlmond
URxvt.foreground: Black
URxvt.color7: Gray90
URxvt.colorBD: Blue
URxvt.cursorColor: Black
URxvt.throughColor: #8080f0
URxvt.highlightColor: Grey70
URxvt.underlineColor: yellow
URxvt.geometry: 80x40
URxvt.loginShell: true
URxvt.meta: ignore
URxvt.utmpInhibit: true
URxvt.scrollstyle: plain
URxvt.scrollBar: true
URxxt.scrollTtyOutput: true
URxvt.secondaryScroll: true
URxvt.savesLines: 1000
URxvt.jumpScroll: true
URxvt.secondaryScreen: true
URxvt.scrollWithBuffer: true
I have a button on my Gnome panel that runs urxvt -geometry
80x75 -sb -si -sl 1000 to open a nice tall window.
Last summer, below, I described how I set up a wireless network. I've set it up again, but this time better. Last time, I tried to get WPA encryption going, but it turns out that the wi device driver, which my wifi card used, doesn't support WPA. All I could use is WEP, which isn't good enough.
After some looking around I found that the ath driver is the best supported for doing WPA. It handles cards that have a chipset from Atheros, and this Atheros customer products database gave me a list of matching cards. I got a D-Link DWL-G650 and it's worked out perfectly.
I set up up my Linksys WAP54G access point as I'd done before. I held down the little button on the back to reset it, then plugged it into my laptop. It defaults to thinking it's at 192.168.1.245, so I ran
ifconfig xl0 down ifconfig xl0 192.168.1.1 ifconfig xl0 up add route default 192.168.1.1
Now the laptop and the access point were on their own little network. I hit http://192.168.1.245/ with Firefox and changed the admin password and the SSID, and disabled a couple of things. All of a sudden, it stopped working. The access point's web server was inaccessible or would return a zero-byte file. Had I broken something? This seemed like bad luck. Someone recommended I try a different browser, so I went back at it with Opera, and this time everything worked perfectly. The WAP54G must be buggy if it won't work with Firefox.
Using Opera, I changed the network settings and set up WPA-PSK with a long and difficult passphrase. I moved the access point and plugged it into the switch in the middle of my home network, and went back to the laptop to get it going.
I had to add some options to wiggins's kernel and recompile it. (These may be in GENERIC now, but I've carried my kernel config over through a few upgrades.)
# Wireless NIC cards device wlan # 802.11 support device wlan_wep device wlan_ccmp device wlan_tkip device wlan_xauth device ath # Atheros PCI/cardbus NICs device ath_hal # Atheros HAL (Hardware Access Layer) device ath_rate_sample # SampleRate TX rate control for ath
Getting the laptop on the network turned out to be pretty easy, thanks
to the documentation and some examples I found around the web. I put
this in /etc/rc.conf:
ifconfig_ath0="WPA inet 10.1.0.7 netmask 255.255.255.0 media autoselect channel 6 ssid MY_SSID"
That WPA at the start clues it in to use
/etc/wpa_supplicant.conf, which looks like this:
ctrl_interface=/var/run/wpa_supplicant
ctrl_interface_group=wheel
network={
ssid="MY_SSID"
bssid=01:02:03:0D:0E:0F
scan_ssid=1
key_mgmt=WPA-PSK
psk="mylongandverycomplicatedpassphrase"
}
The BSSID is the MAC address of the access point.
After a reboot, everything worked. While I was getting it going, I
occassionally ran wpa_supplicant -i ath0 -c
/etc/wpa_supplicant.conf to get WPA going by hand. Also,
/etc/rc.d/netif start (or stop) is handy
while testing, though I found it didn't set up a default route so I
had to run route add default 10.1.0.1 by hand after
stopping and starting.
Last week I listened to Dan Chudnov's talk ZeroConfMetaOpenSearch. In it, while he's getting people in the room to join a wireless network he set up on an access point he'd brought in, he notices that someone who's running Linux on his laptop is visible on the list of people sharing music with iTunes over the network. Dan mentions the guy's running RhythmBox. I thought that was interesting, so I fired up RhythmBox, which was already on my machine because it's part of GNOME, and had a look.
In order to have a real look, and share music between my desktop box (sherlock) and my laptop (wiggins) I had to get avahi going, and running it by hand wasn't working. So I thought, what the heck, I'll run full-out GNOME for a bit just to see this music sharing feature. Then I'll go back to fvwm.
By the end of the evening I'd permanently switched to GNOME. Whatever objections I had to it (the look, that it's a GUI and I only wanted a window manager, whatever) are gone. I like GNOME a lot and I'm sticking with it. The FreeBSD GNOME Project's web site was helpful in getting me moved over.
Here's a screenshot of what my desktop looks like now (110 KB PNG). I've set things up so that:
The default configuration was pretty fine as it was, and I just made minor adjustments so that I didn't have to use any new keystrokes. One very nice thing is that when I move work one workspace to another, it remembers what window was in focus. Alt-tab is workspace-specific, so I can Ctrl-arrow to move to another workspace and if I'm not on the window I want (I probably am, because it's the one I was on when I left) I can Alt-tab to another. It's much easier now to get around without a mouse.
I added this to my kernel configuration because something or other I tried complained about not having it:
# For GNOME. device pass # Passthrough device (direct SCSI access)
And I added this to /etc/rc.conf:
gnome_enable="YES" dbus_enable="YES" polkitd_enable="YES"
I think that just gnome_enable should have done it, but I
got warnings about the other two, so I specified them fully and the
warnings stopped.
I changed exec fvwm2 to exec gnome-session
in my ~/.xinitrc file, but I'm not sure if that's
necessary given that I turn on full GNOMEness in
rc.conf. I haven't bothered to find out.
So far I'm very pleased and I wish I'd gone over to GNOME before.
Over the last couple of months I had some annoying problems with my high-speed ADSL connection to the Internet. When I pumped any serious number of packets through it the line would die, and it required many attempts to reconnect. It turned out that my modem was on the fritz and sending bad signals (or something) and when I got a new modem, everything was fine again.
Along the way, though, I decided it was a good time to replace my old gateway/firewall box, mrs-hudson. mrs-hudson was running FreeBSD-4.12, which is years old, and for security I wanted to run the STABLE branch, now at 6.2. mrs-hudson had a 100 MHz Pentium chip and 16 megabytes of RAM, which isn't much, but was enough for what it was doing. There wasn't enough disk space to compile a new world, though, so I'd had to leave it alone.
For $60 I bought a used Pentium III 600 MHz box with 128 megs of RAM.
I got two used 3Com ethernet cards; there's a network interface built
in, so it has three now. I got it on the network and installed
6.2-STABLE, then used cvsup to download the latest source
updates. I built a new world and kernel and moved to 6.2-RELEASE.
Everything went fine. I copied over my name server files,
/etc/ppp/ppp.conf, and a handful of other files, and I
was ready. I set it in place and now I have a new gateway and
firewall machine: widmerpool. (I've changed my naming scheme from Sir
Arthur Conan Doyle's Sherlock Holmes stories to Anthony Powell's
A Dance to the Music of Time.)
One network interface is connected to my internal LAN, one plugs into my ADSL modem, and one is unused. (I may use it for a public wifi access point.) I use Network Address Translation on widmerpool, and it handles all the work of managing how sherlock and wiggins go out to the Internet.
Here's the relevant but from /etc/rc.conf:
# ----- NETWORKING network_interfaces="xl0 xl1 xl2" hostname="widmerpool.miskatonic.org" # xl0 is empty # xl1 is ADSL # xl2 is the LAN ifconfig_xl2="inet 10.1.0.1 netmask 255.255.255.0" gateway_enable="YES" # ----- PPP ppp_enable="YES" ppp_mode="ddial" ppp_nat="YES" ppp_profile="profilename"
I use 10.1.0.* as the IP
numbers on my LAN. gateway_enable magically makes
widermpool act as a gateway, and ppp_enable means the
Internet connection comes up automatically. ppp_nat turns
on network address translation which also just works magically.
Once I had that in place I was back where I'd started: a gateway/firewall box that handled my Internet connection and had no open ports. I moved my printer over to widmerpool, and running lpd opened a port. I didn't want that. I either had to figure out how to have lpd listen only on the internal interface (dull) or finally do what I'd wanted to try for a years: set up a proper firewall. Fun!
Two things were particularly helpful in setting up the firewall:
Three things gave me trouble:
tun0, not
xl2, which is the interface plugged into the ADSL modem.
Here's what I ended up with. First, what I have in
/etc/rc.conf on widmerpool, the gateway box:
# ----- Firewall ipfilter_enable="YES" ipfilter_rules="/etc/ipf.rules" ipmon_enable="YES" ipmon_flags="-D /var/log/ipfilter.log" ipnat_enable="YES" ipnat_rules="/etc/ipnat.rules"
Now that I look at that again I notice that
/var/log/ipfilter.log doesn't get rotated properly, but I
set /etc/syslog.conf to log everything into
/var/log/all.log, so I can see all the ipmon
output there. (ipmon is what reports on what the firewall
does.)
Here are my ipf.rules and ipnat.rules files. Originally I had the IPFilter rules set to block everything in and out and then allow particular things, but I changed it so that everything is allowed out by default. If I were running a firewall at an office I'd change that and be more restrictive, but for my personal use, it's fine. Everything inbound is blocked by default.
I keep an eye on the logs to see what's going on and what gets blocked. So far everything has been fine.
My time zone moved from Eastern Standard Time (UTC-5) to Eastern Daylight Time (UTC-4) today, and it happened three weeks earlier than it did before. sherlock and wiggins didn't notice it. I thought they would, but for whatever reason they didn't, so I had to do it by hand.
As root, I ran tzsetup and then adjkerntz
-s. That did the job.
I've been using F-Spot to
organize my digital photos. It's in the ports at
graphics/f-spot, and I prefer it to kphotoalbum,
which for KDE users. F-Spot is for GNOME people, and I still
use fvwm, but I can run GNOME apps. In this case, however,
there's a twist.
After using portupgrade to move F-Spot up to the latest version
(0.3.2) and also upgrade mono and some related ports, F-Spot
stopped working. I got this error: "Unable to open the system message
bus." I submitted a bug report: ports/108843:
f-spot 0.3.2,1 fails with bus-related error. After a bit of back
and forth with some helpful people, it emerged that because fvwm
doesn't use dbus, I need to launch F-Spot specially:
dbus-launch f-spot.
I installed the www/linux-opera and
www/linux-flashplugin ports and Flash worked right away.
I'd tried to get the www/flashplugin-mozilla thing
working, which lets the Linux version of the Flash player work with a
FreeBSD-compiled browser, but had no luck. I could have installed
www/linux-firefox but went with Opera for variety.
Macromedia makes available a Linux Flash player, but not a FreeBSD
one. Luckily FreeBSD can run Linux binaries.
Recently I upgraded LaTeX (actually, teTeX) from the ports. Tonight I was updating something important and when I went to generate a fresh PDF, I got this:
Warning: pdflatex (file cmr7.pfb): cannot open Type 1 font file for reading Warning: pdflatex (file cmmi10.pfb): cannot open Type 1 font file for reading Warning: pdflatex (file cmbx10.pfb): cannot open Type 1 font file for reading Warning: pdflatex (file cmti10.pfb): cannot open Type 1 font file for reading Warning: pdflatex (file cmsy10.pfb): cannot open Type 1 font file for reading Warning: pdflatex (file cmcsc10.pfb): cannot open Type 1 font file for reading Warning: pdflatex (file cmtt10.pfb): cannot open Type 1 font file for reading Warning: pdflatex (file cmr10.pfb): cannot open Type 1 font file for reading
Because it couldn't find these fonts, it used some horrible other font and everything looked wrong. It seemed like an odd problem to have, because these are the Computer Modern fonts, designed by the great Knuth himself, and they're a basic part of LaTeX. However, it was easy to fix: I ran updmap ("update font map files for TeX output drivers") and then all was well.
I have a Linksys WAP54G wireless access point, and after some fiddling I got wiggins hooked up to it. Here's how.
The WAP54G comes from the shop thinking that it's 192.168.1.245. I ran a cable from my laptop to its back port. To make my laptop talk to it, I ran
ifconfig xl0 down ifconfig xl0 192.168.1.1 ifconfig xl0 up add route default 192.168.1.1
Now I could hit the access point with my web browser at
http://192.168.1.245/. I logged in and first thing changed the admin
password. To make it sit properly on my network, I set it so it had
the IP number 10.1.0.101 and used 10.1.0.1 as the gateway. Once I'd
saved that I couldn't talk to it any more from wiggins without
reconfiguring. I commented out the ifconfig line from
/etc/rc.conf, so it would boot up without doing any
network stuff, and then I plugged the access point into my switch so I
could control it from sherlock, my desktop machine.
Now I could hit the access point at http://10.1.0.101/ from sherlock, and I had to configure it and then make the laptop match. I changed the SSID and disabled the broadcasting of same. I turned on MAC filtering and made it so that only my wireless ethernet cards would be allowed to connect. This didn't work. The access point kept deleting them from the list. It turned out the firmware on my device was at version 3.01, and there's a bug in it that causes just this problem. I hit Linksys's web site, downloaded the new firmware (version 3.04), discovered I had to unzip it before uploading it to the access point, and got it installed. It reported an error, but everything seems fine. After that, I set up the MAC filtering and the settings stayed put.
Now I went over to wiggins and tried to get it on the network. It worked. The final step was to get WEP working. I set it up on the access point and then added it to wiggins's configuration, and it worked. I know WEP can be broken easily, but it helps against casual sniffing. I use ssh for all traffic back and forth and usually leave the access point turned off when I'm not using it, anyway.
I run this script when I want to get wiggins going with wifi:
#!/bin/sh echo "Configuring wi0 for home ..." ifconfig wi0 10.1.0.7 ifconfig wi0 media autoselect ifconfig wi0 channel 6 ifconfig wi0 ssid MY_SSID ifconfig wi0 authmode open ifconfig wi0 wepmode on ifconfig wi0 weptxkey 1 ifconfig wi0 wepkey 0x_MY_WEP_KEY_IN_HEX ifconfig wi0 up
I could put all this into an ifconfig line in
/etc/rc.conf, and tried, but it didn't work. Until I get
that figured out, running this script is no big chore.
Not having to use an ethernet cable is great, but now my bandwidth has gone from about five megabits to 300 kilobits per second.
A couple of weeks ago I came home and discovered that sherlock, my main desktop computer, wasn't working. The monitor said there was no video signal and hitting keys didn't do anything, but the power was on and lights were lit. I turned it off and on and the power came back on but nothing happened. That was not good. Computers are supposed to beep when you turn them on. Sherlock, however, was no POST, no beep, no video. Some browsing around the web led me to think my motherboard was dead. That was not good.
Before I took the box into the shop I wanted to make sure my hard
drives were OK. I have two in sherlock: the main one and then another
with /usr/ports and a bunch of MP3s. I keep everything
backed up: every week I copy everything I don't want to lose to
another machine, I burn it to CD-ROM, and I copy it to an external
USB drive which I
lock in a safe, encase in cement, and then store in a nuclear bomb
shelter in an undisclosed location.
I pulled out the last CD-ROM and checked all the files, and they were
all fine except for the backup of my home directory. About half the
files were there, then I got an error: 'x' header failed?!
This can't happen. That was not good. Where were all the files?
What had happened?
It turns out that the SessionSaver Firefox extension had a file in it dated 1901, and that tar has a bug and can't handle very old files. Here's how to recreate it:
$ touch -t 191007011800 foo $ ls -l total 0 -rw-r--r-- 1 buff wheel 0 Jul 1 1910 foo $ tar czvf foo.tar.gz foo a fooarchive_write_pax_header: 'x' header failed?! This can't happen.
All of my home directory backups since January, when I moved up to FreeBSD 6.0, were incomplete! Horrible news. I had made the fatal error of making backups but then not testing them. If my hard drives were dead then I would lose work.
Luckily, they weren't dead, and I got them working easily. The only
other computer I had was a laptop, wiggins, and I couldn't take the
drives out of sherlock and put them into wiggins. However, I could
use an external USB enclosure! I took the main drive out of sherlock,
hooked it up in the enclosure, and plugged it in. I knew that
sherlock's /etc/fstab had these two lines:
/dev/ad0s1b none swap sw 0 0 /dev/ad0s2a / ufs rw 1 1
On wiggins I mounted /dev/da0s2a (switching devices from
ad, the "generic ATA/ATAPI disk controller driver," to
da, the "SCSI Direct Access device driver" that all USB
devices show up as) and voila, there was sherlock's hard drive, in
perfect shape. I touched the 1901 file, backed everything up
again, and I was all set. If the new motherboard meant I had to
reinstall everything, I was OK.
As it turned out, I was lucky. Not only were both hard drives undamaged, when the new motherboard and CPU were installed, sherlock booted up normally, saw the drives, and everything was fine. The new motherboard even had the same chipset on the on-board video card, but I reinstalled X just to be on the safe side.
It's not perfect, though. I do have problems with my USB drives, so I filed a problem report. However, both the problem with my MP3 player and the problem with my external drive can be worked around, so I have some minor inconveniences but no major ones, and no loss of data.
I was having annoying problems with strangely-named directories in my
~/.mozilla/ directory where Firefox kept its files. They
had unprintable characters in their names, and every time I upgraded
Firefox, new ones would be made. I had to make a profile for myself
and reset Firefox each time to use that profile. The directory names
looked like this:
drwx------ 2 buff buff 512 Jan 7 00:53 ?D&(???/ drwx------ 2 buff buff 512 Jan 1 14:41 ?D&( ??/ drwx------ 2 buff buff 512 Jan 1 14:41 ?D&(0??/ drwx------ 2 buff buff 512 Jan 8 14:16 ?D&(@??/ drwx------ 2 buff buff 512 Jan 1 14:40 ?D&(@??/ drwx------ 2 buff buff 512 Jan 1 14:42 ?D&(H??/ drwx------ 2 buff buff 512 Jan 8 14:18 ?D&(P??/ drwx------ 2 buff buff 512 Jan 1 14:41 ?D&(P??/ drwx------ 2 buff buff 512 Jan 9 23:36 ?D&(X??/ drwx------ 2 buff buff 512 Jan 9 23:37 ?D&(h??/ drwx------ 2 buff buff 512 Jan 22 17:29 ?D&(h??/ drwx------ 2 buff buff 512 Jan 15 10:58 ?D&(x??/ drwx------ 2 buff buff 512 Jan 11 21:52 ?D&(???/
It turned out all this was caused by me having built a debugging
version of Firefox. I removed /var/db/ports/firefox/,
rebuilt the port, did not turn on debugging, and now all is well.
I saw some discussion on the freebsd-stable mailing list about how stable 6.0 is and when it's safe to move up from 5.x. I'd thought it was the same deal as with previous versions and regular folks should wait a while, until 6.2 or 6.3, before upgrading. It's not, though; 5.x will probably only have one more release and the recommendation was that people move up to 6.0. It's stable and it's easy to upgrade without doing a fresh install.
I read the instructions and did an upgrade on a machine I don't care so much about, and it worked fine, so I did it on two others and had no problems. Here's what I did:
stable-supfile to get the new source
tree:
*default host=cvsup9.freebsd.org *default base=/var/db *default prefix=/usr #*default release=cvs tag=RELENG_5 *default release=cvs tag=RELENG_6 *default delete use-rel-suffix *default compress src-all
cvsup -L 2 -g
stable-supfile
/usr/src/UPDATING
options COMPAT_FREEBSD5 to my kernel
configuration file (though I forgot to do this on one machine and
nothing bad happened).
/usr/src/UPDATING:
cd /usr/src make buildworld make buildkernel KERNCONF=SHERLOCK make installkernel KERNCONF=SHERLOCK reboot [come up in single user mode and proceed if nothing goes wrong] fsck -p mount -u / mount -a cd /usr/src mergemaster -p make installworld mergemaster -i reboot [breathe sigh of relief that it worked]
lib*,
gnome*, all the multimedia ports, X Windows, all
Perl/Python/PHP ports, and more.
Everything is working fine. I've had no trouble at all. It was
nerve-wracking, and mergemaster had a lot of files to handle
and I had to merge /etc/passwd and a couple of others by
hand, but it all went smoothly and was infinitely preferable to doing
a fresh install and then restoring everything from backups and the
ports.
My old Logitech mouse, just a regular three-button mouse with a ball inside, was giving me trouble. Sometimes it would stop moving left and right and only move up and down. I cleaned it, but the problem kept on. I decided to pick up an optical mouse, one of the kind that has a little light shining out the bottom and uses reflections of that light to figure out how it's being moved. There's nothing to get gunked up.
I bought a cheap Labtech mouse. It didn't work right. The pointer
was jumping all around it was somehow causing the computer to think
keys had been pressed or something had happened that made it paste
text in all over the place. That was no good. Luckily, someone else
had had this exact problem and
posted to a mailing list about how he'd fixed it. I did what he
did and now the mouse works: I added this line to
/boot/device.hints:
hint.psm.0.flags="0x200"
Now I see this when I boot:
psm0: <PS/2 Mouse> flags 0x200 irq 12 on atkbdc0 psm0: [GIANT-LOCKED] psm0: model Generic PS/2 mouse, device ID 0
I upgraded Firefox to 1.5 (well, 1.5_2,1) tonight.
~/.mozilla/ because there was no way
to make it see any dot directories when I wanted to choose the
directory where the profile is. The "All Files" button didn't work.
about:config and set
middlemouse.contentLoadURL to true. Somehow this got set
to false.
sysctl hw.snd.pcm0.vchans=4 sysctl hw.snd.maxautovchans=4
Now I more than one thing can play sounds at the same time, and I don't get complaints about not the audio device being locked by another process.
I had a weird problem but I found the cause. The other night I
noticed that the GNOME stuff had all been moved up to a newer release
and there were dozens of ports to reinstall. I ran the upgrade script
(mentioned in /usr/ports/UPGRADING) and had only one
small problem where I had to delete an existing port so that the
script could install it on its own.
When I rebooted, there was trouble. After a few seconds of using an xterm, every keypress would cause a horizontal line to appear on the screen. If I held a key down, the flickering lines would scroll down down the screen, sort of like a TV with horizontal hold problems. I couldn't figure it out. How could upgrading GNOME cause this?
I degaussed my monitor, I turned it off an on, I rebuilt X, I
rebooted, I unplugged and replugged my keyboard, I checked my X log, I
built a new xorg.conf, I used twm as my window
manager instead of fvwm, I rebuilt the world and upgraded the
entire operating system. Nothing worked. I searched for mentions of
problems with my graphics card (Silicon Integrated Systems [SiS]
unknown chipset (0x6330) rev 0, an onboard card using up 64
megs of memory), and found some mentions (see this page by the
driver writer) that convinced me it was the video card.
I couldn't think of anything else it could be, so I went around
checking prices on cards. I thought I'd try one last thing —
swapping in an old monitor to make sure my monitor wasn't the trouble
— when I noticed a line in my .xinitrc I'd skimmed
over before:
xscreensaver &
I had to isolate every possible cause, so I commented that out and started up X again. No lines! No problems! xscreensaver depends on GNOME, so it was rebuilt with everything else, and something is now tickling a bug somewhere. I'll report it. Since I was only using xscreensaver to black my screen, now I just let my monitor's automatic power-saving handle things. All is well.
I bought a new USB flash drive, but I got an error when I tried to mount it. It came with a DOS file system installed on it, so I was mystified. A quick newfs_msdos /dev/da0 reformatted the drive and then it mounted without any trouble.
My Epson Stylus C60 inkjet printer stopped working a little while ago.
Really, it probably does still work, but the new ink cartridge I'd
bought dried up or gave up after five days and ten pages, and I said
the hell with it. Inkjet cartridges are an enormous ripoff, and I
decided to free myself from the international inkjet conspiracy and
buy a laser printer. After some looking around at low-end models by
Dell, Brother, Samsung, HP, and Lexmark, I decided that the Brother
HL-2040 looked good. It's not listed on linuxprinting.org as being
supported, but this
message from a mailing list archive said it did work with CUPS and
the hl1250 driver. Since other similar Brothers work
without any trouble, I was pretty sure this would too.
When I got the printer home and plugged it in, all seemed well. I
used the USB
connection and it was recognized (because I have USB stuff and ulpt
enabled in my kernel):
ulpt0: Brother HL-2040 series, rev 2.00/1.00, addr 2, iclass 7/1 ulpt0: using bi-directional mode
I tried using apsfilter, but after the configuration, a test page was requested but never emerged from the printer. Was it a problem with apsfilter? I didn't know. Now, there are about 10n (for some n>2) Linux people using CUPS, and any web search about Unix/Linux printing leads into people talking about it, PPD files, and such. All this was a mystery to me, but if that's what it took to get the printer working, I'd try it.
I installed cups from the ports, ran cupsd, and
(the port doesn't tell me this, but it'd be nice if it did) hit
http://localhost:631/admin/ to use the web interface. It
asked me for a username and password. You're supposed to log in as
root, but I disabled the authentication in the config file (whose
location I forget) so I didn't have to bother.
cups doesn't know about the hl1250 driver out of
the box, so I had to download
a PPD file and copy it into
/usr/local/share/cups/model/ (setting perms so it was 644
root:wheel) so that I could use it. Note: you need to run
killall -HUP cupsd for it to see the new file. I ran
through adding a printer (I told it to use a local USB connection, which it saw)
and then tried to print a test page, but again, nothing came out. It
didn't say why.
Now I was getting worried that the printer wouldn't work. The printer has a parallel port on it, though, so I tried that, and it worked right away with apsfilter. Phew! I don't know what stopped it from working over USB, but I don't really care. I was just glad it worked and that I didn't have to do all that CUPS stuff.
apsfilter did everything for me after I ran
/usr/local/share/apsfilter/SETUP and answered a few
questions. The hl1250 driver is listed under
"Miscellaneous Other Drivers." My /etc/printcap now
says:
lp|hl1250;r=300x300;q=medium;c=mono;p=letter;m=auto:\
:lp=/dev/lpt0:\
:if=/usr/local/etc/apsfilter/basedir/bin/apsfilter:\
:sd=/var/spool/lpd/lp:\
:lf=/var/spool/lpd/lp/log:\
:af=/var/spool/lpd/lp/acct:\
:mx#0:\
:sh:
I want to turn some old radio shows on cassette tape into MP3s. A bit of poking around made me reckon that Audacity would be just what I needed to handle the actual recording, and indeed that turned out to be so, but I had a bit of trouble actually getting the sound to go to the right spot. My sound card was working well for output,I knew that because I could play MP3s with xmms, but I had trouble with the input.
I ran a line from the back of my tape player (the REC OUT; don't use the headphone jack because you'll get a hum) to the microphone input of the sound card (I needed a wire with two male RCA jacks on one end and one male 1/8" plug on the other) but Audacity wasn't hearing anything.
I fixed it with the mixer utility. Running it with no arguments showed something like this:
Mixer vol is currently set to 75:75 Mixer pcm is currently set to 75:75 Mixer speaker is currently set to 75:75 Mixer line is currently set to 75:75 Mixer mic is currently set to 0:0 Mixer cd is currently set to 75:75 Mixer rec is currently set to 0:0 Mixer line1 is currently set to 75:75 Recording source: cd
Notice that the mic volume is set to 0, and the source is cd. I ran
mixer mic 90 and mixer +rec mic and then
got:
Mixer vol is currently set to 75:75 Mixer pcm is currently set to 75:75 Mixer speaker is currently set to 75:75 Mixer line is currently set to 75:75 Mixer mic is currently set to 90:90 Mixer cd is currently set to 75:75 Mixer rec is currently set to 0:0 Mixer line1 is currently set to 75:75 Recording source: mic
Audacity then added the microphone to its list of inputs and pressing the record button worked. I could grab some audio then export it as an MP3 or Ogg file. The shows I had were split across two sides of a tape, so I hit pause while I turned it over, then edited out what empty space remained. Editing is really easy. It takes quite a while to export an hour of audio as an MP3, but the result turned out well.
Below, on 12 January 2005, I mention that on my new 5.x system the
/dev/cdrom symlink to /dev/acd0 would
disappear when I rebooted. Roland Smith sent me
some tips about how to stop this. As he says on his web page,
"FreeBSD 5.x has a dynamic /dev filesystem. This means
that entries in /dev are created or removed depending on
the available devices." The file to control this is
/etc/devfs.conf. It contains these lines:
# Commonly used by many ports #link acd0 cdrom
I uncommented the second. There's a lot more that be controlled in that config file, but this is all I need right now.
I used XML::RSS to generate an RSS feed. It was quite easy, but there were three things I want to remember.
First, the actual RSS file
should have the proper
MIME
type. For my RSS 1.0 feed,
which is in the RDF format, I added this to Apache's
httpd.conf: AddType application/rdf+xml .rdf
Second, the date format I chose from the specification to use
in the Dublin Core metadata should look like
2005-04-17T16:11:26-04:00, which a colon in the middle of
the time zone offset. I had to splice that in because
Date::Manip only seemed to have it without the colon.
Third, this will help with autodiscovery:
<link rel="alternate" type="application/rdf+xml"
title="Some Name Here (RSS 1.0)" href="/url/to/feed.rdf" />
Though how legit application/rdf+xml is in the link tag,
I don't know. But it works in Firefox.
xargs -I
In a misguided attempt at cleaning out my Firefox cache at the
command line, I deleted at directory named Cache and
somehow wiped out all my customizations, including my bookmarks and
extensions. I updated the Firefox port to 1.0.1, and got set
to install everything all over again. Here are the extensions I use:
The hard part came when I had to restore my bookmarks.
Firefox installed into directories with strange
characters in their names, so I can't just cd into them.
ls ~/.mozilla/ looks like this:
-rw-r--r-- 1 buff buff 335 Mar 2 14:45 appreg drwxr-xr-x 2 buff buff 512 Mar 2 15:01 ?x#(0X?/ drwxr-xr-x 2 buff buff 512 Mar 2 14:45 ?x#(P??/ drwxr-xr-x 2 buff buff 512 Mar 2 14:45 ?x#(p??/ drwxr-xr-x 2 buff buff 512 Mar 2 14:45 ?x#(???/ drwxr-xr-x 2 buff buff 512 Mar 2 16:17 ?x#(???/ drwxr-xr-x 2 buff buff 512 Mar 2 14:45 ?x#(???/ drwxr-xr-x 3 buff buff 512 Mar 2 14:45 ?x#(???/
I think it's a character set problem, but I'm not sure what's going
on. Everything works all right for programs, but not for people. I
had to get my backup bookmarks.html into place without
being able to actually move into the proper directory, and I found a
way to do it with find and xargs with the latter's
-I option (notice how find's output show different
directory names):
buff:sherlock:~/.mozilla$ find . | grep bookmarks.html | xargs ls -l -rw-r--r-- 1 buff buff 16513 Mar 2 15:52 ./?x#(???/7vt9f154.default/bookmarks.html buff:sherlock:~/.mozilla$ find . | grep bookmarks.html | xargs -I "%" cp ~/lost-bookmarks.html % buff:sherlock:~/.mozilla$ find . | grep bookmarks.html | xargs ls -l -rw-r--r-- 1 buff buff 54520 Mar 2 16:09 ./?x#(???/7vt9f154.default/bookmarks.html
I did that with Firefox shut down, of course, and when I started it up again, all my bookmarks and feeds were back.
I was talking to some fellow geeks about USB drives, and one had an old 64 meg key drive he generously gave me. He'd been using a half-gig drive for a while and this small one was compartively ancient.
Getting it working on both my laptop (wiggins) and desktop (sherlock) machines was pretty simple. First, I made sure this section was in my kernel config (taken from the GENERIC kernel):
# USB support device uhci # UHCI PCI->USB interface device ohci # OHCI PCI->USB interface device usb # USB Bus (required) #device udbp # USB Double Bulk Pipe devices device ugen # Generic device uhid # "Human Interface Devices" device ukbd # Keyboard device ulpt # Printer device umass # Disks/Mass storage - Requires scbus and da device ums # Mouse device urio # Diamond Rio 500 MP3 player device uscanner # Scanners
The umass device is commented out in GENERIC, but I
uncommented it and added this to bottom of the config:
# Required for USB. device scbus device da
I rebuilt world and kernel and updated everything following the proper procedures. Booting sherlock now shows:
usb0: OHCI version 1.0, legacy support usb0: SMM does not respond, resetting usb0: <SiS 5571 USB controller> on ohci0 usb0: USB revision 1.0 uhub0: SiS OHCI root hub, class 9/0, rev 1.00/1.00, addr 1 uhub0: 3 ports with 3 removable, self powered
wiggins is pretty much the same. I stuck the key drive into the slot to see if it would work, and it did. This appeared in the console window:
umass0: vendor 0x1005 USB Flash Drive, rev 2.00/1.00, addr 2 da0 at umass-sim0 bus 0 target 0 lun 0 da0: <USB Flash Drive 1.12> Removable Direct Access SCSI-0 device da0: 1.000MB/s transfers da0: 61MB (126973 512 byte sectors: 64H 32S/T 61C)
Perfect. But how to get to the files there? I thought the drive had
probably been formatted with an MS-DOS file system, so I put this into
/etc/fstab:
/dev/da0 /usb msdosfs rw,noauto 0 0
I ran mkdir /usb; mount /usb and the drive seemed to be
mounted, but the directory was empty so I couldn't be sure. I copied
a file into it, ran umount /usb, took the drive to the
other computer, mounted it, and saw the file. Excellent!
I can use amd to automatically mount and unmount drives, and if I get that working then I'll make a note here.
This isn't specific to FreeBSD, but I'll mention it anyway. In the database that by personal library catalogue program Mrs. Abbott uses, I have tables called AUTHOR, TITLE, and HASWRITTEN. AUTHOR has author names, TITLE has information about particular books, and HASWRITTEN is a matrix table that ties together AUTHOR.id (which is HASWRITTEN.authorref as a foreign key) and TITLE.id (HASWRITTEN.titleref) to show who wrote what.
I wanted to do a select that would show me all of the writers who are not responsible for any books in my collection, that is, writers whose books I used to have but have since weeded. In MySQL 3, I had to do it the primitive way:
SELECT author.* FROM author LEFT JOIN hasWritten ON author.id=hasWritten.authorRef WHERE hasWritten.authorRef IS NULL;
I'm a bit hazy on left and right and inner and outer joins, but it works. It's easier in MySQL 5, though, because since 4 they've allowed subselects:
SELECT author.*
FROM author
WHERE NOT EXISTS (SELECT *
FROM hasWritten
WHERE hasWritten.authorRef = author.id);
MySQL SQL Syntax and Use explains how to rewrite subselects as joins.
My personal library catalogue system, Mrs. Abbott, uses MySQL.
On my old box I'd been using version 3.something, but they've moved up
to 4.1 as the production release version and are working on 5. With 5
I could have views, triggers, subselects, the whole megillah of what I
was missing in 3, so I went with it. Installing it from the ports was
as easy as usual, but then I couldn't get it started. I ran
/usr/local/etc/rc.d/mysql-server.sh start, but nothing
happened. I ran ls -l /usr/local/bin/mysql* and saw
mysqld_safe, which was mentioned in the boot script, so I had a
look inside.
# mysql.server works by first doing a cd to the base directory and from there # executing mysqld_safe
Aha! It was being very secure. I ran cd /var/db/mysql;
/usr/local/bin/mysqld_safe and it started up. In another
shell, as root I ran mysql -u root -p mysql to get to the
admin interface, and then I just repeated what I'd done on 3 July
2002, below (changing the hostname where necessary) except for one
thing: when it came to granting privileges, I ran
mysql> grant all privileges on library.* to marion@localhost; Query OK, 0 rows affected (0.00 sec) mysql> grant all privileges on library.* to marion@sherlock.miskatonic.org; Query OK, 0 rows affected (0.01 sec)
There were some new privileges added in version 4, and without the ability to lock tables my mysqldump script wouldn't work. There's no reason why the marion user shouldn't have complete rights over the library database, and those lines set all that up. My backup script looks like this:
mysqldump --user=marion --password=hushhush --complete-insert --add-drop-table library > \
$DATE-mysql-library.sql
I've used xine before, but never much liked it for playing
MP3s. The interface is pretty ugly and I never bothered to download
any skins to change it. I tried gxine, which looks much
simpler, and I like it. Both, it turns out, use the
multimedia/win32-codecs port, which lets them show
Windows Media files and all kinds of things I've never been able to
see outside of a Windows box. Installing gxine also added a
bunch of lines to my .mailcap file and now I can see
embedded media in web pages. Nice.
There was one problem, though: some Quicktime videos didn't show any
pictures, just a blue box. The xine FAQ
suggested adding Option "XaaNoOffscreenPixmaps" to the
device section of /etc/X11/xorg.conf, but that didn't
help. The next
question in the FAQ said I should run xine -V XShm
and see if that worked, and it did. This means there is something
buggy with the Xv extension I have, and I have to use another video
plugin thing. I have no idea what the cause of the problem is, but it
all works now.
I found a bug in how gxine is installed from the ports: it
compiles and installs without any trouble, but when run it complains
that it can't find libgnomevfs.so.0 (which is from the
libgnomvfs version 1 port). I submitted a
problem report about this, which got bounced around a couple of
times and is now still open. Installing libgnomevfs from
the ports makes it shut up, and doesn't interfere with GNOME 2.
I was able to burn CDs without any trouble (though my
/dev/cdrom symlink keeps disappearing, which is
annoying). However, when I tried to mount a CD, I got this error:
cd9660: /dev/acd0: Input/output error. I looked around
some mailing list archives and someone suggested ejecting and
reloading the disc, and that did the trick. I'd feared it would be
something much more serious.
The heating problems I mentioned below just got worse, and soon I couldn't keep up a load of 1 for more than five minutes without the machine's temperature alarm going off. Since the load went to 1 if I was listening to music, using Firefox, and doing some other small tasks, it was unusable. I went out and bought a new box:
I bought my first real PC almost exactly ten years ago, a 486 DX2/66 with 16 megs of RAM. It cost $1700 and I had to take out a loan to pay for it. Five years later I got mycroft, a PII-450, and now this. Who knows what I'll have in 2010?
Here are some notes about what happened when I installed 5.3-STABLE and got everything in shape.
At the very start, I got warnings about the geometry of the disk being incorrect. It said it would accept what the BIOS said, so I rebooted and checked it, but when I entered those numbers they were rejected. I figured I'd go on anyway, since FreeBSD was bound to be pretty smart, and see how it worked out. It's fine. I ended up having a problem I thought was related to the geometry, but it wasn't: because of the way I installed the boot manager and arranged the slices, the first couple of times I tried to boot the machine after the install, I tried to boot off the swap partition. There's no way that will ever work. I was being offered a choice between F1 and F2, and when I hit F2 instead, it saw the right boot stuff and found the kernel. The geometry seems fine. The disk is working perfectly, in any case.
The sound card is onboard, built into the motherboard, and figured it'd be some kind of standard type, but wasn't sure. I was a bit nervous about how this and the ethernet and video cards would work, but FreeBSD and X grokked them. To get sound working, first I put
device sound
in my my kernel config, but that didn't do anything. No pcm device showed up on boot. I ran pciconf -lv which showed some interesting stuff, including:
none0@pci0:2:7: class=0x040100 card=0x70121849 chip=0x70121039 rev=0xa0 hdr=0x00
vendor = 'Silicon Integrated Systems (SiS)'
device = 'SiS7012 PCI Audio Accelerator'
class = multimedia
subclass = audio
but after running kldload /boot/kernel/snd_ich.ko it showed this:
pcm0@pci0:2:7: class=0x040100 card=0x70121849 chip=0x70121039 rev=0xa0 hdr=0x00
vendor = 'Silicon Integrated Systems (SiS)'
device = 'SiS7012 PCI Audio Accelerator'
class = multimedia
subclass = audio
How did I know that was the right one to load? I used Google on
"Silicon Integrated Systems" freebsd and that sort of
thing. To get the sound device in the kernel config so it was built
right in, I added
device snd_ich
(which I tried as a guess after just ich didn't work) and
that compiled. Now when I boot, I see
pcm0: <SiS 7012> port 0xd800-0xd87f,0xdc00-0xdcff irq 18 at device 2.7 on pci0 pcm0: [GIANT-LOCKED] pcm0: <Unknown AC97 Codec (id = 0x434d4983)>
Balance control in xmms didn't work, and stuff sounded like it was in mono. I enabled a SuperStereo plugin and now it sounds stereo, though the balance and volume knobs do nothing.
This got a bit annoying. Two parts of the X.Org distribution in the
ports were broken: xorg-clients and
xorg-server. They are crucial! I worked around it by
installing packages: pkg_add -r xorg-clients and pkg_add -r
xorg-server. I had to run pkgdb --fix to fix up some
dependency misunderstandings. GNOME2 was able to go in without any
trouble. Not that I use any of it, but a few applications I like
require it.
When I got X fired up, I found my resolution was 2048*1536 and it refuses to go lower! I want to get it down to 1600*1200. I had to jack my font sizes up to 17.
Emacs wanted libXaw3d.so.8, but
x11-toolkits/Xaw3d was installing
libXaw3d.so.6. There was a bug about this
filed
last November. The fix is to make a symlink so everything's
happy:
bash-3.00# cd /usr/X11R6/lib/ bash-3.00# ls -l libXaw* -rw-r--r-- 1 root wheel 444594 Oct 16 00:31 libXaw.a lrwxr-xr-x 1 root wheel 11 Oct 16 00:31 libXaw.so -> libXaw.so.7 -rwxr-xr-x 1 root wheel 276475 Oct 16 00:31 libXaw.so.6 -rwxr-xr-x 1 root wheel 386642 Oct 16 00:31 libXaw.so.7 -rw-r--r-- 1 root wheel 398424 Jan 9 13:03 libXaw3d.a lrwxr-xr-x 1 root wheel 13 Jan 9 13:03 libXaw3d.so -> libXaw3d.so.7 -rwxr-xr-x 1 root wheel 298262 Jan 9 13:03 libXaw3d.so.7 bash-3.00# ln -s libXaw3d.so.7 libXaw3d.so.8 bash-3.00# ls -l libXaw* -rw-r--r-- 1 root wheel 444594 Oct 16 00:31 libXaw.a lrwxr-xr-x 1 root wheel 11 Oct 16 00:31 libXaw.so -> libXaw.so.7 -rwxr-xr-x 1 root wheel 276475 Oct 16 00:31 libXaw.so.6 -rwxr-xr-x 1 root wheel 386642 Oct 16 00:31 libXaw.so.7 -rw-r--r-- 1 root wheel 398424 Jan 9 13:03 libXaw3d.a lrwxr-xr-x 1 root wheel 13 Jan 9 13:03 libXaw3d.so -> libXaw3d.so.7 -rwxr-xr-x 1 root wheel 298262 Jan 9 13:03 libXaw3d.so.7 lrwxr-xr-x 1 root wheel 13 Jan 9 13:16 libXaw3d.so.8 -> libXaw3d.so.7
I added
proc /proc procfs rw 0 0
to /etc/fstab so truss would work.
I got X going, but my mouse started acting up. All of a sudden the
pointer would appear at the edge of the screen, menus would start
popping up, and then huge chunks of text would get dumped into an
xterm. This was really annoying. I looked around and found it in the
FAQ: Why
does my PS/2 mouse misbehave under X?. The FAQ said I should
"disable the synchronization check code by setting the driver flags
for the PS/2 mouse driver to 0x100. Enter UserConfig by giving the -c
option at the boot prompt," that is, boot: -c, but I
couldn't get to that prompt. boot -c just booted up normally,
and I didn't get the UserConfig menu.
Checking some mailing lists told me that if I couldn't send the signal
this way, I could add it to the device hints read at boot. I added
this to /boot/device.hints:
hint.psm.0.flags="0x100"
That had no effect. Didn't fix the problem at all. I plugged in a
new mouse, rebooted, and it acted crazy even in console mode. I
removed the line from device.hints, rebooted, and then
the new mouse worked perfectly. Who'd have thought a mouse would
cause so much trouble? Why is one different from another? Who knows.
I followed the advice in the handbook about adding disks and used /stand/sysinstall. I couldn't mount my second drive from mycroft as it was (some error about an incorrect superblock), so I had to copy over the data first and then reformat it. This involved a fair bit of moving the drive from one box to another. I tried using one of those neat rounded IDE cables, and had no trouble removing the old cable, installing the new one, and hooking up the two drives. However, when I closed the case it must have pushed the cable in such a way that something stopped working, because the second drive wasn't seen on boot. I had to cannibilize my old box and take the old ribbon cable that connected the two CD drives. It's not as pretty inside, but no-one will ever see it.
mycroft is now five-and-a-half years old and has a PII-450 CPU, which means it's so old that a used computer store would just have it in a big stack of junk in the back. It's served me well, though, aside from the time the hard drive up and died. Lately the CPU fan started to get louder, and louder, and then it became noisy, and then I knew I had some trouble because the CPU temperature warning beep would go off if the load hit 1 for more than a few minutes.
As I've mentioned before, I know little about hardware and am especially leery of fooling around with anything important, but this required attention. I opened up the case (all work inside was done with the power cable unplugged, of course) and sprayed some compressed air all around. A fair bit of dust blew out, but the fan didn't quiet down. I unscrewed the fan and gave it a good blast, but that didn't help either. I ran the computer without the fan, but it overheated fairly quickly. I got a real fan and aimed it at the open case, but that didn't help.
Finally I decided to put the fan back and make plans to take the box somewhere so I could either get it fixed (fans are very cheap) or get a new CPU installed. I was surprised to see that the screws holding the fan in place didn't fit into special holes, they just went right into the spaces between slats in the heatsink. I assume they are just the right size to push the slats slightly out of the way so they can get a good grip. It's a little offset from where it had been before (not that it matters) and since then I've had no cooling problems. I haven't built a new world or reinstalled Mozilla — the ultimate tests — but I can update lots of ports without any trouble. I'm out of immediate danger.
I had some audio files that had been broken up into small chunks, for example a half-hour talk cut up into ten-minute sections. sox does a great job of concatenating the files. I just ran sox short1.mp3 short2.mp3 short3.mp3 full-file.mp3 and it did all perfectly. Very handy.
So I wanted to install the latest version of GeneWeb, a genealogy program written in the obscure language Caml. It's not in the ports, so I got the source from the download page. The latest version is 4.09. I untarred it, ran make, and got:
buff:mycroft:/usr/local/src/geneweb-4.09$ make cd wserver; make opt camlp4r pa_ifdef.cmo -DUNIX -o wserver.ppi wserver.mli ocamlc.opt -warn-error A -I `camlp4 -where` -c -intf wserver.ppi File "wserver.mli", line 24, characters 15-46: The type constructor format expects 3 argument(s), but is here applied to 4 argument(s) *** Error code 2 Stop in /usr/local/src/geneweb-4.09/wserver. *** Error code 1 Stop in /usr/local/src/geneweb-4.09.
I have no idea what that means. There's no mention of it on GeneWeb's site. They have a mailing list on Yahoo! Groups, so I joined. I should have joined a while ago. Now, I joined by sending in an e-mail subscribe message, because I don't want a Yahoo! Groups account. The Yahoo! Groups web site is horrible. But to read the mailing list's archives, you have to have an account. Ugh! So I used the BugMeNot extension to Firefox and got a password, signed in, joined the list, and immediately discovered that the problem comes from a particular release of Caml and the solution is to get the latest source from CVS. Fine. But why isn't this mentioned anywhere public? Why not roll a new release so the problem never arises? Once I'd got the new source and compiled it, I had my fake account leave the group. An annoying waste of time.
For the last few rebuilds I've followed the super-safe recommended
procedure, as described in /usr/src/UPDATING.
cd /usr/src make buildworld make buildkernel KERNCONF=MYCROFT make installkernel KERNCONF=MYCROFT shutdown -r now
Wwhile it's rebooting I hit a letter or the shift key to stop the usual boot process. Then I need to tell it to boot into single user mode, and then I mount the disks and finish the install. If there are any problems with the new kernel, I'll see them now and I can reboot again, run boot kernel.OLD, and recover.
boot -s fsck -p mount -u / mount -a cd /usr/src adjkerntz -i make installworld mergemaster -vC reboot
Below, on 12 January 2002, I described how I set up my cheap Epson Stylus C60 (the printer was cheap--the ink cartridges sure aren't). Since then my box died and was rebuilt, and in the meantime I'd just left the printer hooked up to my Windows box where I was doing my word processing. Now that I'm using LaTeX, though, I need the printer back on my main machine so I can use good old lpr to print stuff.
I installed apsfilter from the ports, and ran
/usr/local/share/apsfilter/SETUP. Telling it to use
/dev/lpt0 and 8.5" x 11" paper was easy, but I had a
hard time finding the right driver. It couldn't find the drivers from
Epson, nor the ghostscript drivers, nor the GIMP
drivers, and I was starting to wonder what the hell it had been doing
while it was dropping all those files on my drive. Then I found the
stcany driver (I forget in which menu, and I don't want to run
the setup script again in case it overwrites something) which is for
"Epson Style Color (any model)." Perfect.
I ran a test page and it looked good. I'd called the printer
epson again, but to make it the default I changed the
name in /etc/printcap:
lp|stcany.upp;r=300x300;q=medium;c=full;p=letter;m=auto:\
:lp=/dev/lpt0:\
:if=/usr/local/etc/apsfilter/basedir/bin/apsfilter:\
:sd=/var/spool/lpd/epson:\
:lf=/var/spool/lpd/epson/log:\
:af=/var/spool/lpd/epson/acct:\
:mx#0:\
:sh:
(I think that next time I could just install apsfilter, put
that in my printcap, make the directories mentioned,
and it'd work.) I didn't do anything to
/usr/local/etc/apsfilter/apsfilterrc. I ran lpd,
ran lpc and brought the printer up and down and up again,
printed out a Postscript file I'd made with LaTeX, and
everything worked nicely. Adding lpd_enable="YES" to
/etc/rc.conf wrapped it all up. Next, I'll see about
getting it working under samba, which I couldn't do before.
Also, I'd like to find an ink monitor.
I can't remember what finally made me want to install LaTeX, but I'm glad I did. It's great and I'm going to stop using OpenOffice for writing, except for when I need to wrangle a Microsoft file format. I installed teTeX from the ports, and it came with all sorts of packages and tools that make life easy.
Since then I've heard mention in news of some other packages, and I
found them at CTAN. The memoir
class is the biggest one I've added, and it's very nice. I don't
know of a fancy LaTeX package manager that installs things from the
ports or somewhere else, so I did it by hand, following instructions
from the FAQ on installing
a new package and where
to put your files.
Two useful references:
I was getting some annoying errors when I aimed Emacs at Perl
files. It wasn't finding cperl-mode.el properly. I got it
fixed by changing my .emacs file. Now I use
defalias to say that when I'm dealing with Perl, I want
cperl-mode:
;; cperl-mode (defalias 'perl-mode 'cperl-mode) (setq cperl-indent-level 4) (setq cperl-continued-statement-offset 3) (setq cperl-close-paren-offset -2) (setq cperl-info-on-command-no-prompt t)
That stopped it from complaining. Then Jeremy
Zawodny put me hep to Tramp. It's like ange-ftp, but
you can use scp so it's secure. I installed it from the ports
(net/tramp), then added this to my .emacs:
;; TRAMP is like ange-ftp, but better. (require 'tramp) (setq tramp-default-method "scp")
Now, how do I use it? Well, in my ~/.ssh/config I have
an alias that makes it easy to ssh into the Unix box where I
host my mail and web site:
Host shell Hostname shell.example.com User buff
That lets me run ssh shell and make a connection. And now, in
Emacs, I can open up /shell:file.txt and, after
much output explaining what it's doing, it'll open up a buffer where I
can edit a file on the remote machine. The network becomes completely
transparent, the way it should. Very nice.
sox said it couldn't decode MP3s, so I recompiled it with
WITH_LAME, WITH_MAD, and WITH_VORBIS all defined. Then it said it
wouldn't work on this huge 60-minute long track. Turns out
xmms will do the trick. I went to Options | Preferences
| Audio I/O Plugins and set the Output Plugin to "Disk Writer
Plugin 1.2.8 [libdisk_writer.so]" which came from who knows where. I
hit play, and it started going silently through the file at about ten
times normal speed. I ended up with a big WAV file in my home
directory, where I'd told it to write.
The WAV file couldn't be burned as an audio disc because it was the wrong format, though. But now sox got friendly again:
sox -c 2 -r 44000 big-file.wav burn-to-cd.wav
That turned it into the kind of thing you can make an audio disc from, but, while fooling around with all this burning, suddenly I started getting:
burncd -v -f /dev/cdrw -d audio burn-to-cd.wav fixate burncd: ioctl(CDRIOCWRITESPEED): Input/output error
I think I broke my CD burner.
I read in news that it's possible to pretend that ATAPI drives are SCSI devices. I've never used SCSI, and don't really care about it, except that I had the idea perhaps this would be a way I could burn proper music CDs.
First, I had to add some lines to my kernel config:
device atapicam device scbus device cd device pass
I recompiled the kernel and installed it, rebooted, and voila, there they were:
cd0 at ata1 bus 0 target 0 lun 0 cd0: <CREATIVE CD4830E 1.01> Removable CD-ROM SCSI-0 device cd0: 16.000MB/s transfers cd0: cd present [6018 x 2048 byte records] Mounting root from ufs:/dev/ad0s1a cd1 at ata1 bus 0 target 1 lun 0 cd1: <CDWriter IDE128 1.04> Removable CD-ROM SCSI-0 device cd1: 16.000MB/s transfers cd1: Attempt to query device size failed: NOT READY, Reserved ASC/ASCQ pair
camcontrol and cdrecord will tell more:
$ camcontrol devlist
<CREATIVE CD4830E 1.01> at scbus1 target 0 lun 0 (pass0,cd0)
<CDWriter IDE128 1.04> at scbus1 target 1 lun 0 (pass1,cd1)
$ cdrecord -scanbus
[ ...]
1,0,0 100) 'CREATIVE' 'CD4830E ' '1.01' Removable CD-ROM
1,1,0 101) 'CDWriter' 'IDE128 ' '1.04' Removable CD-ROM
[ ... ]
First, I had to get the raw files off a music CD: cdparanoia --output-raw -B. Then, to burn them: cdrecord -v speed=4 dev=1,1,0 -audio *raw. That took about 20 minutes (one quarter of 80 minutes, a full disc), but in the end I had the same problem as before: the song start times get messed up when I play the disc in a proper player, and none of the songs start at the start. I think the problem is not with how I'm writing the discs, but with my CD burner.
Recently I used portupgrade on GNOME things, and something went a bit funny. After a reboot, I fired up X again and of course started Galeon. I got an error:
GConf Error: Failed to contact configuration server (a likely cause of this is that you have an existing configuration server (gconfd) running, but it isn't reachable from here - if you're logged in from two machines at once, you may need to enable TCP networking for ORBit)
It turned out I had both gconf and gconf2 installed, and 2 was running and had something locked while 1 was trying to run, or something like that. I can't keep all that GNOME stuff straight. I uninstalled gconf, thinking perhaps I wouldn't need it any more, but I did. After uninstalling and reinstalling both, Galeon worked again. The reason remains a mystery.
I installed the www/plugger and www/flashplugin-mozilla
ports today, and they work just fine in Galeon.
plugger, which has a ton of dependencies so it can handle a lot
of different file types, dropped a file in
/usr/X11R6/lib/browser_plugins, while
flashplugin-mozilla laid one on me in
/usr/X11R6/lib/mozilla/plugins. PDFs show up right in the browser, not in a
popped-off helper application, and I can see all those Flash things
people talk about. It all seems fairly solid, too--no crashes yet. I
may try getting the RealPlayer to work as plugin. I don't see
what the big deal is about plugins--I don't mind handing something
over to a helper app--but some people make it really hard for things
to work any other way.
I got GNOME 2.2 installed on my work machine easily. I followed the
instructions
and ran pkg_delete on the listed ports, did a make
deinstall on x11/gnome, did a make install on
x11/gnome2 (and deleted some more packages when it told
me to) and all was well.
I read a number of weblogs, and everyone's talking about RSS, which is a way of using XML to distribute the weblog entries (or news article from a newspaper, or what have you). I can use an RSS aggregator to go out, read the feeds, collect everything, and show it to me all on one screen. It saves me from having to visit 25 different weblogs myself. I don't really mind that, but I wanted to try out an aggregator to see what it's like.
I found a couple in the ports: straw and krss. krss requires KDE, which I don't want to install it just for one little program. straw requires GNOME, which I already have installed because I use galeon and Mozilla. straw it would be, but it turned it out needed a recent version of some Python library or other, which required GNOME 2. I had GNOME 1. I had to upgrade. That's where the trouble started.
First I went into /var/db/pkg and ran pkg_delete -f
gnome*. Then I reckoned there must be a better way to do it. The
FreeBSD GNOME
2.2 FAQ recommended something, but I read the answer to the next
question, which said I should run pkg_delete -rf pkgconfig\*.
I did, and when it was over half of X was gone, along with
Emacs, galeon, fvwm, and a bunch of other things
I use every day. Luckily X and fvwm kept on working,
and I was able to install GNOME 2 from the ports and then catch up on
the other stuff. I even installed a bunch of extra font packages
while I was at it. When it was all over, everything was back to
normal, everything worked, and when I went to run straw, I got
...
Traceback (most recent call last):
File "/usr/X11R6/bin/straw", line 31, in ?
import straw
File "/usr/local/lib/python2.2/site-packages/PIL/__init__.py", line 16, in ?
# ;-)
File "src/lib/Main.py", line 13, in ?
ImportError: No module named ui
There's a report about this in their bug database, but the answer didn't fix it for me. Elsewhere I've seen people saw they just can't get straw working at all. I fiddled a bit, then said the hell with it.
Compiling is slow and buggy on the laptop, so I made some packages on
mycroft using pkg_create. It's easy to do with files in
/var/db/pkg:
$ cd /var/db/pkg/fontconfig-2.2.0/ $ pkg_create -v -c +COMMENT -d +DESC -f +CONTENTS -m +MTREE_DIRS ~/fontconfig-2.2.0.tgz Creating package /home/buff/fontconfig-2.2.0.tgz Creating gzip'd tar ball in '/home/buff/fontconfig-2.2.0.tgz'
Copy that to the other machine and it's perfect for handing to pkg_add.
It took an evening to get wi-fi going on my laptop, simpson. It wouln't have taken so long if I hadn't had an IRQ conflict. I found the solution hidden in a FreeBSD mailing list's archives. Why is PC hardware so annoying? Why should I have to worry about such a stupid thing? If I ever buy a new laptop I'll get one from Apple with OS X.
For now, though, I have an IBM Thinkpad i1411 with a PII-300 Mhz CPU and 32 megs of RAM. (This is so wimpy I doubt they even sell them in used computer stores. On top of that, something's flaky with the hardware (perhaps memory) because it sometimes gives segmentation faults (the old signal 11 error) when compiling.) I also have a Linksys WPC11-CA 802.11b card, which FreeBSD knows under the more general term "Linksys Instant Wireless WPC11."
To my great delight, the laptop recognized the wi-fi card when it booted up, and knew it needed the wi driver. The last time I tried to get a PCMCIA card working with FreeBSD, I had to fool around with config files for days. That's another thing about open-source operating systems that must keep away casual users.
wi0 at port 0x240-0x27f irq 9 slot 0 on pccard0 wi0: 802.11 address: 00:03:19:as:2f:e1 wi0: using RF:PRISM3(PCMCIA) wi0: Intersil Firmware: Primary 1.01.00, Station 1.04.02
Anyway, I looked at the pccard entries in
/etc/defaults/rc.conf and customized
/etc/rc.conf so that pccard_enable was on and other
pccard options knew about my network, and set the ifconfig entry for
wi0 to DHCP. Didn't work. I got errors about how it "failed to
allocate 1594 bytes on NIC." Searching the web told me this was
because of an IRQ
conflict. grep -i irq /var/run/dmesg.boot showed two things
battling for IRQ 9: the
PCI bus and the
wi-fi card. The wi-fi card had to come out on the losing end of that
fight. I needed to make it use another IRQ. I cleaned up /etc/rc.conf
leaving only pccard_enable set, so that the other stuff wouldn't get
in my way, and searched the web.
I found the answer in some FreeBSD mailing list archives (after
searching for "freebsd irq conflict" and things like that), provided
by guru Warner Losh. I had to set hw.pcic.intr_path=1 in
/boot/loader.conf. This is mentioned
in the release notes and described in pcic(4):
hw.pcic.intr_path
Should PCI pcic devices route interrupts via ISA or PCI. A value of 1 means route via ISA. A value of 2 means route via PCI. This is ignored for the ISA device. Many older laptops do not have PCI BIOS implementations that FreeBSD can use to route interrupts properly. These laptops may need to set this to 1 and hw.pcic.irq to a value (or 0 for polling). The default is 2.
With that in /boot/loader.conf, I rebooted and now the
wi-fi card came up on device 3. That was a bit better, but it was
also taken--however, it was also listed as a valid IRQ number in
/etc/defaults/pccard.conf. All I had to do was tell the
system that number wasn't free. I created
/etc/pccard.conf and put in this after looking at
dmesg to find out what numbers were unused:
irq 11 13
I rebooted, and this time the card came up on IRQ 11! No conflicts. Perfect!
After all that, I had a working wi-fi card in my laptop. Now I had to
get it on the network. First, of course, I reconfigured my wireless
access point so it would allow the connection. To keep things simple,
I decided not to trigger any events out of
/etc/pccard.conf when the card was inserted or removed.
I made a shell script to do everything:
#!/bin/sh echo "Configuring wi0 ..." # Channel 11 wicontrol -f 11 # Infrastructure mode wicontrol -p 1 # SSID wicontrol -n "my-ssid" # Turn up WEP (key in hex) wicontrol -e 1 wicontrol -k 0x12345678901234567890123456 echo "Getting IP number with DHCP ..." dhclient wi0
All those wicontrol commands can be put on one line, but this
way shows how the details of the configuration are built up one by
one. I run that by hand when I boot up and it almost always connects
to the network. A few times it didn't, but I rebooted and it worked.
I'm not sure what the problem was. If I customize
/etc/pccard.conf to do all this for me, I'll write it up here.
The wicontrol (8) man page and these web pages (and some others I can't find now) were very helpful to me while I figured all this out.
An old IBM Thinkpad laptop fell into my hands and I wanted to run FreeBSD on it and get wi-fi working. First, I had to install FreeBSD, and along the way I learned how to build my own FreeBSD release. I had an old 4.something-RELEASE CD around, but I wanted to use the latest version. For some bizarre reason I couldn't download any ISO images--they'd go for a while, then freeze, no matter which mirror I tried. I'm not sure if it's my connection to my ISP or what. After some frustration I decided to avoid all the trouble by building my own release, burning it to CD and booting and installing from it. When you use an open source operating system, this is on the road to starting your own distribution, but don't worry, I'll never go that far.
I read the release (7) man page and looked at makefiles, but still had trouble. I asked for help on the freebsd-stable mailing list, and Peter Jeremy and Igor Pokrovsky cleared things up. Later I got confused about the CD images, and Bruce Mah explained. After a few attempts, I got it working:
vn module into the kernel by running
kldload vn.ko in /modules/.
/etc/make.conf (see 3 April 2003), because the system's
regular version of Perl had to be installed. cvs-supfile to download the
CVS repository
into /home/ncvs/.
/disk2/src/ and /disk2/release/.
Then I ran this:
cd /disk2 cvs -d /home/ncvs/ -R co -rRELENG_4 -P src cd src make buildworld cd release make release CHROOTDIR=/disk2/release \ BUILDNAME=4.8-PRERELEASE-MISKATONIC CVSROOT=/home/ncvs \ RELEASETAG=RELENG_4 NOPORTS=YES MAKE_ISOS=YES NODOC=YES
That makes a whole new source tree from the CVS repository, just like in
/usr/src, then does the familiar make buildworld,
followed with a make release with a few options so the ports
and docs are left out but ISO images are made. This takes a
while and uses up a lot of disk space.
I ended up with disc2 and miniinst ISO
images in /disk2/release/R/cdrom. Where was
disc1? The maintainers make it by hand--it's just the
miniinst disc with packages added. All I needed was
4.8-PRERELEASE-MISKATONIC-i386-miniinst.iso. I burned a
copy and made floppy images from kern.flp and
mfsroot.flp in
/disk2/release/R/cdrom/disc1/floppies. It installed on
the laptop without any problems.
I upgraded X to version 4.3, using the ports, and had a couple
of small problems. The compile failed twice, when something about
either popt or Xft was wrong and a dependency failed.
It was pretty easy see from the errors where the problem was, and
since reinstalling usually fixes things like that, that's what I did,
and that's what happened. Once the new X was actually running,
I couldn't use galeon, because something it needed relied on a
libintl.so.2, but I had libintl.so.4. Now,
I could have made a symlink, but that just masked the problem. So I
did without galeon for a day while I recompiled everything with
portupgrade -frv gettext (which is the port that makes the
libintl files). When that was done, galeon
worked, but the gettext port disappeared. I don't know where
it went, but it often caused me troubles when using
portupgrade, so I'm not sorry to see the back of it.
I had a bit of Perl work I wanted to do which required some new
modules. While installing from CPAN, I realized I was tired of the
old Perl that FreeBSD uses, so I installed version 5.8 from the
ports, then ran use.perl port to make the new version the one
the system would use. It added a few lines to
/etc/make.conf.
# -- use.perl generated deltas -- # # Created: Mon Mar 10 21:07:02 2003 # Setting to use base perl from ports: PERL_VER=5.8.0 PERL_VERSION=5.8.0 PERL_ARCH=mach NOPERL=yo NO_PERL=yo NO_PERL_WRAPPER=yo
I had to reinstall all the modules I had, which was pretty easy with
portupgrade -frv p5* in /var/db/pkg (though
actually I did them one by one to watch what happened). Then I fired
up the CPAN module (with perl -MCPAN -e 'shell') and ran
install Bundle::CPAN. It installed more modules, and it noted
them in /var/db/pkg:
bsdpan-Archive-Tar-0.23, bsdpan-CPAN-1.70
and so on. perldoc BSDPAN explains a bit about what this is,
including that it "registers Perl modules with FreeBSD package
database." Very nice.
favicon.ico
I managed to make myself one of those favicon.ico icon
things. I read these instructions for
Unix users, which left me needing only an actual graphic to use.
My graphic manipulation powers stop at resizing GIFs, but I managed to
do up a 16x16 PNG
in the GIMP. I cranked it up to 1600%
magnification, put an M on, then used the pinpoint editing tool to
change the colour on pixels. It worked. Depending on your browser,
you may see a red M on a white background in your location bar or in
your bookmark list if you bookmark this. Galeon shows them in
its tabs. Looks like Miskatonic University Press finally has a logo.
In other news, a while ago I changed the VNC server on wiggins, my Windows laptop. The client on mycroft stopped working. It would core dump as soon as the screen was drawn. I guess the client and server have to match up, which had never occurred to me. Yesterday I installed TightVNC on both boxes, and everything's working. It's faster, too, but not as fast as using RDP, which is super-fast.
.fvwm2rc again
I made some updates to my fvwm2 set-up. I realized that the
FvwmWinList thing, which makes a list of windows in use,
was just taking up screen real estate. I never used it, except to
marvel at how many xterms I had open. I also realized that
never in my life had I ever used two of the three buttons I placed on
the top right corner of all windows. They minimized and maximized the
window, but I never do that. So I took them out, and while I was at
it I changed the top left button to what seems to be the default, a
bar, and I changed some colours. Now my root window is
SteelBlue4. Here's a screenshot, and here's my new
.fvwm2rc. If you compare to a screenshot from six years
ago, you'll see things aren't much different. My screen has
better resolution so I can fit more on, there's no more bar along the
bottom, and the button bar has moved from the bottom right to top
right, but it's still basically the same. My window manager looks are
getting simpler as the years pass. This is the cleanest yet.
After
reading about the acronym tag and seeing how useful
it is, I ran through this page and put in definitions for most of the
acronyms.
I was updating some ports the other day and had problems compiling
xmcd and some X stuff and I thought it was a bug with
open-motif, but it turned out I was missing a
hosts.def file that imake depended on. A header
file tried to include it but it wasn't there. I grepped through
/var/db/pkg/*/+CONTENTS to see where it came from, and
reinstalled the port. That fixed it. Odd.
Nothing exciting has happened FreeBSD-wise in the last half of 2002, but I hope to get a wireless network going soon and if I do I will write it up.
I was pretty pleased with the way my hard drive installation had gone, and since the jinx seemed to be gone, I figured now was the time to get a CD burner. I thought they were expensive but I found one for $60, which I thought was a deal. I'd be getting an ATAPI burner, so I had to get something on the list of supported drives, but that list is out of date. I got a newer model BTE, and it turned out to be OK.
My CD-ROM
drive was in the top third of a metal case at the top of the computer
case, and the other two slots had some kind of metal gizmo in them.
It looked like something should fold down and support the drive, but I
ended up ripping the whole thing out and that solved the problem. I
set both drives to cable select and put the CD burner in as slave, so
it's my secondary slave now. It shows as /dev/acd1c, and
I made a /dev/cdrw symlink to that.
Using an ATAPI burner means I have to use burncd, which comes with the base system. I installed mkisofs as well. These pages were very helpful:
To make a data CD, first I had to make an ISO 9660 image.
mkisofs -v -J -L -R -V "20020718 Backups" -o backup.iso BACKUPS
I ran that in the directory above where my BACKUPS/
directory is, and it made an ISO 9660 image containing all the files
inside. The options make it verbose, turn on some thing that allow
better filenames, and set a label. mkisofs has tons of
options. Once the image is made, I wanted to make sure it was all
right. There's no equivalent to tar's tvf, so you
have to do this, which is cool:
# vnconfig /dev/vn0c backup.iso # mount_cd9660 /dev/vn0c /mnt # ls -l /mnt ... # umount /mnt # vnconfig -u /dev/vn0c
You mount the ISO image as a fake file system and then check it out. Wild! I've never done that before. Everything looked fine, so it was time to burn.
burncd -e -v -f /dev/cdrw data backup.iso fixate
That went fairly quickly, and voila, my backup was done. I put the disc my regular CD drive, mounted it, compared the files, gunzipped and untarred them, and they were perfect. After some more experimenting--this was on a CD-RW, so I could blank it and try over--and I burned a CD-R and it was fine. To blank a CD-RW, you do this:
burncd -v -f /dev/cdrw blank
I did a lot of that while testing. I hosed one CD-RW, too, thought I don't know why. The drive wouldn't see it. I threw it out. To make an ISO image of an existing CD, I did this:
dd if=/dev/cdrom of=quicktax.iso bs=2048
You need to specify bs or else it won't work.
Everything's worked nicely, except that I can't make an audio CD. I
took some WAV files and did what burncd's man page said to do,
but all I got was a CD that my computer thought had one song on it,
which went backwards in time, and which my CD player wouldn't
register. If I get it working, I'll post notes here. It's nice to be
able to make your own mix CDs, but mostly I want this for backups (and
MP3s).
One other thing: my NFS
mounts weren't working, but the problem was that I forgot that while
mount points on the same disk have to be listed on the same lines in
/etc/exports, different disks have to be on different
lines. My /usr/ports is actually
/disk2/ports, and since it's a different disk (and
symlinks can't be listed) I put it on a separate line and now it's
fine.
I noticed these lines in my syslog:
Jul 18 09:34:56 mrs-hudson sshd[15667]: scanned from 218.4.74.53 with SSH-1.0-SSH_Version_Mapper. Don't panic. Jul 18 09:34:56 mrs-hudson sshd[15666]: Did not receive identification string from 218.4.74.53.
They come from ScanSSH. My sshd is right up to date, so I'm not worried.
My ADSL problems were caused by a faulty DSLAM port at the central office. It was fixed and my line's working perfectly.
I got samba working by copying what I did on 27 November 2001,
below. Worked the first time, except I forgot that later I'd set
read only to no so I could write to the network-mounted
drive.
Now that I've got a high-speed connection (which Bell says they fixed
again, and it's working well tonight), I want to get into some of that
Internet audio before it goes away or I have to pay. I browsed around
and got confused by all the pls and m3u
files and MIME types. RealAudio worked OK, but
what did I need to do to get streaming MP3s working? After a bit of
looking around I added some lines to my .mailcap and
.mime.types files. To my .mailcap:
# Streaming MP3s audio/mpg;xmms %s audio/mpeg;xmms %s audio/x-mpeg;xmms %s audio/mpegurl;xmms %s audio/mpeg-url;xmms %s audio/x-mpegurl;xmms %s audio/x-mpeg-url;xmms %s # MP3 URL audio/m3u;xmms %s audio/x-m3u;xmms %s # Shoutcast playlists audio/scpls;xmms %s audio/x-scpls;xmms %s
And to my .mime.types:
audio/mpg mp3 mp2 mpega MP3 audio/mpeg mp3 mp2 mpega MP3 audio/x-mpeg mp3 mp2 mpega MP3 # MP3 URL audio/mpegurl m3u audio/mpeg-url m3u audio/x-mpegurl m3u audio/x-mpeg-url m3u audio/m3u m3u audio/x-m3u m3u # Shoutcast playlists audio/x-scpls pls audio/scpls pls
The various RealAudio entries I left alone. Lynx can handle all these kinds of files properly now, and just hands them off to xmms (creating an instance if it's not already running). In Galeon I had to add some MIME types and handlers in by hand, which seems stupid. I still can't get the PLS files working in it, but it seems like most people use M3U. Live365 works, and they've got lots of different kinds of music, and lots of different kinds of annoying pop-ups.
Next on the list was MySQL and Mrs. Abbott, my library catalogue program. I installed MySQL from the ports. Again it said
PLEASE REMEMBER TO SET A PASSWORD FOR THE MySQL root USER ! This is done with: /usr/local/bin/mysqladmin -u root -p password 'new-password' /usr/local/bin/mysqladmin -u root -h mycroft -p password 'new-password' See the manual for more instructions.
I ran mysql:
/usr/local/etc/rc.d/mysql-server.sh start
And again, as on 2 November 2000, I couldn't set the password the way they recommended. Later I thought perhaps I should have done
mysql -u root -p
and gone from there, but by that time it was too late. I'd done it as I'd done it before:
root:mycroft:/usr/local/etc/apache2$ mysql -u root -p mysql
Enter password: HIT ENTER
Reading table information for completion of table and column names
You can turn off this feature to get a quicker startup with -A
Welcome to the MySQL monitor. Commands end with ; or \g.
Your MySQL connection id is 5 to server version: 3.23.51
Type 'help;' or '\h' for help. Type '\c' to clear the buffer.
mysql> SELECT Host, User, Password FROM user WHERE User = 'root';
+-----------+------+----------+
| Host | User | Password |
+-----------+------+----------+
| localhost | root | |
| mycroft | root | |
+-----------+------+----------+
2 rows in set (0.00 sec)
mysql> SET PASSWORD FOR root@mycroft = PASSWORD('hushhush');
Query OK, 0 rows affected (0.01 sec)
mysql> SET PASSWORD FOR root@localhost = PASSWORD('hushhush');
Query OK, 0 rows affected (0.00 sec)
mysql> SELECT Host, User, Password FROM user WHERE User = 'root';
+-----------+------+------------------+
| Host | User | Password |
+-----------+------+------------------+
| localhost | root | 2d4al9dbwer1e97b |
| mycroft | root | 2d4al9dbwer1e97b |
+-----------+------+------------------+
2 rows in set (0.00 sec)
Looking good. MySQL root passwords set. Next up, create the library
database. I had a slighty out of date dump I wanted to restore. I
had to create a library database and a user named
marion.
mysql> create database library;
Query OK, 1 row affected (0.00 sec)
mysql> grant select, insert, delete, update on library.* to marion@localhost;
Query OK, 0 rows affected (0.00 sec)
mysql> SET PASSWORD FOR marion@localhost = PASSWORD('ontheqt');
Query OK, 0 rows affected (0.00 sec)
mysql> grant select, insert, delete, update on library.* to marion@mycroft;
Query OK, 0 rows affected (0.00 sec)
mysql> SET PASSWORD FOR marion@mycroft = PASSWORD('ontheqt');
Query OK, 0 rows affected (0.01 sec)
mysql> SELECT Host, User, Password FROM user;
+-----------+--------+------------------+
| Host | User | Password |
+-----------+--------+------------------+
| localhost | root | 2d4al9dbwer1e97b |
| mycroft | root | 2d4al9dbwer1e97b |
| localhost | | |
| mycroft | | |
| localhost | marion | aaaaf2fe0c123456 |
| mycroft | marion | aaaaf2fe0c123456 |
+-----------+--------+------------------+
6 rows in set (0.00 sec)
mysql> delete from user where user = "";
Query OK, 2 rows affected (0.00 sec)
mysql> SELECT Host, User, Password FROM user;
+-----------+--------+------------------+
| Host | User | Password |
+-----------+--------+------------------+
| localhost | root | 2d4al9dbwer1e97b |
| mycroft | root | 2d4al9dbwer1e97b |
| localhost | marion | aaaaf2fe0c123456 |
| mycroft | marion | aaaaf2fe0c123456 |
+-----------+--------+------------------+
4 rows in set (0.00 sec)
Now I had root and marion set up. I found that to login as marion I had to have the mycroft host set, so I deleted the localhost one.
mysql> delete from user where user = "marion" and host = "localhost"; Query OK, 1 row affected (0.00 sec) mysql> SELECT Host, User, Password FROM user; +-----------+--------+------------------+ | Host | User | Password | +-----------+--------+------------------+ | localhost | root | 2d4al9dbwer1e97b | | mycroft | root | 2d4al9dbwer1e97b | | mycroft | marion | aaaaf2fe0c123456 | +-----------+--------+------------------+ 3 rows in set (0.00 sec)
But then I couldn't login, so I redid the two commands above to create marion@mycroft. As you can see, I'm a bit hazy on some of these details, but I get there in the end. Now I could run
mysql -u marion -p library
and login to the library database. The last thing was to restore it from the backup. As before, I had to run this as root:
mysql -u root -p library < /tmp/library-20020530.sql
And it was all back. It was a simple matter to get the old Mrs. Abbott Perl CGIs working under Apache with the new directory structure. The MySQL port even installed the Perl DBI modules I needed.
I got Apache going tonight. Usually I install it my own way, but I
decided this time I'd do it from the ports. I installed it, and
edited /usr/local/etc/apache2/httpd.conf. I made myself
the ServerAdmin, set the AddHandler directive for CGIs, and then
decided I should just make all my hosts virtual ones. Also, the host
library needed to have index.cgi work if I didn't want to
use index.html. After a bit of experimenting and a
glance at the docs, I stuck this at the bottom:
NameVirtualHost 10.1.0.3
<VirtualHost 10.1.0.3>
ServerAdmin wtd@pobox.com
DocumentRoot /usr/local/www/server/mycroft/htdocs
ServerName mycroft.miskatonic.org
ServerAlias mycroft
ErrorLog /usr/local/www/logs/mycroft_error.log
CustomLog /usr/local/www/logs/mycroft_access_log combined
ScriptAlias /cgi-bin/ /usr/local/www/server/mycroft/cgi-bin/
<Directory /usr/local/www/server/mycroft/htdocs>
Options Indexes FollowSymLinks ExecCGI
</Directory>
</VirtualHost>
<VirtualHost 10.1.0.3>
ServerAdmin wtd@pobox.com
DocumentRoot /usr/local/www/server/library/htdocs
ServerName library.miskatonic.org
ServerAlias library
ErrorLog /usr/local/www/logs/library_error.log
CustomLog /usr/local/www/logs/library_access_log combined
ScriptAlias /cgi-bin/ /usr/local/www/server/library/cgi-bin/
<Directory /usr/local/www/server/library/htdocs>
Options Indexes FollowSymLinks ExecCGI
</Directory>
</VirtualHost>
/usr/local/www/server/ will be where I house the files
for the sites. mycroft is the working version of this site, and where
I experiment, and library is for my library catalogue. I made those
directories, ran apachectl start, and it worked for mycroft.
Bingo! I made library point to the same IP number as mycroft, and it
all seemed to work, though I got confused for a few minutes because
Galeon was caching old results and it looked like
http://library/ was showing the same home page as
http://mycroft/. There's no "real" host, and I've never
set up a server that way before. It's sensible, though, and works
nicely.
Happy Canada Day.
I got an ADSL connection to E-Gate, which is where I read my mail
and host this site. They shipped me a GVC
BB0040 ADSL modem. I plugged it in, and after a couple of days
the synch light stopped blinking and went solid. I tried to get the
connection up, but the light went back to blinking. Nothing every
goes smoothly. E-gate dealt with Bell and the line was fixed. I got
the connection up very easily. First, I added this to
/etc/ppp/ppp.conf on mrs-hudson:
egate: set device PPPoE:ste0 set authname myusername set authkey mypassword set dial set login add default HISADDR
ste0 is the ethernet card into which the modem is
plugged. In /etc/rc.conf I commented out the
old PPP stuff and added new lines for this modem:
# PPP stuff ppp_enable="YES" # PPPoE ppp_mode="ddial" ppp_nat="YES" ppp_profile="egate" # Regular analog dial-up #ppp_mode="auto" #ppp_nat="YES" #ppp_profile="netcom"
The ddial mode means that if the connection dies, the
daemon will bring it back. I tried ppp -ddial -nat egate
on the command line, and it worked, so I rebooted to make sure it was
all good. Everything on my end was fine.
There was some kind of problem with the line: every two minutes or so the modem would lose synch. The connection would freeze, the light on the modem would blink, then go solid, and my connection would come back. The PPP daemon never noticed anything wrong. It was damned annoying, but tonight I haven't had any problems, so E-Gate's problem report must have been noticed by Bell and fixed. I'm very happy with the speed tonight.
Came home last Thursday and found mycroft had crashed! The hard drive was dead: when I rebooted, it would make clicking noises but never find any files or be able to boot. Argh!
I went out and bought two new Maxtor 40 gig drives. I put them in but the system wouldn't recognize them. The best I could do was to get one to show up as a primary slave. I don't like fooling with hardware, and all the fiddling and recabling was making me a bit nervous. I found the specs for the drive on Maxtor's web site, and noticed they require the newer 80-wire IDE cables. I was using the 40-wire kind, which was all the old drives needed. This page on pcguide.com was very helpful--that's a great site for learning about PC hardware. It's thorough and well-written. I went out and bought some 80-wire cables, but I still couldn't get the drives working.
I'd just assumed my BIOS could support drives that large, but it turns out they can't. Thank heavens for Google's Usenet archive. (I have an ASUS P5A Super7 motherboard, and there's nothing on ASUS's web site about it any more. I won't buy from them again.) I had two choices: upgrade the BIOS, or buy smaller drives. There's no way I was going to fool with the BIOS, and I just wanted the damned thing to work, and fast, so I took the drives back and got two 20 giggers. I took them away, tested them, one worked, the other was broken! Argh! I had to go all the way back and get a replacement, but finally I got it working. I didn't have enough cable to make each drive a master, so I put them both on the primary IDE controller as master and slave, and left my CD-ROM drive as secondary master. I set both drives to "cable select" and it all just worked.
To install the OS, I
burned a CD of the new 4.6-RELEASE and set my BIOS so it could boot off
CD. I stuck the CD in, booted up, and it worked! I didn't need the
two boot floppies like I've always used before. Very nice. I
downloaded cvsup and got my ports and source right up to date,
tweaked the kernel config and added sound support, made world and all
was looking good.
I'd bought two new D-Link NICs (or ethernet cards, as I still call them) to put
into mrs-hudson. One would replace my old NIC for my local network, the other would be
for ADSL, and my old 10 megabit card would wait until I
got a wireless network going. I had some trouble figuring out just
which driver the DFE-530TXS uses, so I just enabled every
D-Link-related driver, installed a new kernel, and waited. It saw
them as ste devices, so I redid the kernel with just that
and it was all set.
Along the way, I learned that there's no way to get into a Compaq Deskpro 2000's BIOS unless you have a special Compaq thing installed (which presupposes that you use Windows). If you have it, then F10 on boot will let you edit the BIOS settings. If you don't have it, you're out of luck. That's really annoying.
Back on mycroft, I symlinked /dev/cdrom -> /dev/acd0c and
got sound going with sh MAKEDEV snd1 and all was right
with the world. I just had a lot of downloading to do, and believe
me, on a 28.8 Kbps modem, it takes a long time to download
XFree86, GNOME, Emacs and all the rest
Conclusions: the entire innards of a PC are stupid. Different cables, different screws, the chances of shorting something, it's nuts. All cabling that's more complicated than a phone cord or an RCA jack is crap. I should be able to buy a disk cube that has some plugs in one end and when I plug it into a slot (into which it fits only one way) in my computer, it just works.
Further to the note on 7 April 2002: I upgraded to Xfree86 4.2
again, and this time it worked. I'd seen the answer go by on the
freebsd-stable mailing list: turns out all I needed to do
was add this line to /etc/X11/XF86config in the "Device"
section:
Option "composite_sync" "Off"
And further to the note of 17 November 2001, it turns out that for the
last month and a half I was making a bit of an ass of myself. My
twice-hourly automated fetchmail download of mail from a
POP server had stopped
working (fetchmail had died and left
~/.fetchmail.pid sitting around, which it uses as a lock
file) but I didn't notice, so I kept sending out mail from the address
but never got any replies. They were accumulating on the POP server. I thought everyone
was ignoring me. Eventually I noticed, and I added some
error-checking to the script:
if [ $? -ne 1 ]; then echo "Fetchmail error downloading FIS mail ... please investigate" | \ mail -s "FIS fetchmail error" wtd@pobox.com fi
Never forget: always check for errors!
A while ago I was upgrading all my ports and found that SWIG-devel ("a tool for automatically generating the wrapper code needed to link collections of functions written in C/C++ with interpreted scripting languages") now required PHP. I don't use it and I didn't want it installed, especially because it in turns requires Apache, which I do use, but installed my own way and not from the port.
I tried some options in /usr/local/etc/pkgtools.conf, but
nothing worked, so I mailed the maintainer. He responded very quickly
and said I should set the environment variable SWIG_LANGUAGES, which
is set in the Makefile thusly:
SWIG_LANGUAGES?= guile perl php4 python ruby tcl
I set it to guile perl python ruby tcl and it installed
just how I wished. I put this into /etc/make.conf so I
won't have to worry about it again, but I'll keep an eye out in case
new languages are added. Now I know what the ? means in
ports makefiles: if it's not set, use this value.
Last week I decided to upgrade XFree86 to version 4.2. I used
pkg_delete -f to get rid of the old version, then did a
make install on the new one. I've had very little
trouble with X in the past and expected none now, but this time it
didn't work. When I started X, it came up normally, then went black
as though I'd killed the session. Then it came back on, then off
... then on ... then off and on. Sometimes fast, sometimes slow. I
didn't know what the hell was happening, but I knew a guy who knew a
guy who made his monitor smoke because he got his XF86Config settings
wrong, so I killed X pronto. I looked around www.xfree86.org and couldn't see
any such problem mentioned, but when I asked on
freebsd-stable, two people who also have Mach 64 video
cards said the same thing had happened to them (one on FreeBSD, one on
Linux). They'd gone back to XFree86 4.1, so I figured I'd
better too.
To do this, I decided to cvsup into the past. I've seen people talk about this but never tried it. First, I had to see when the port had gone to version 4.2, so I checked the ports cvsweb. Revision 1.105, the last while it was 4.1, was dated Mon Jan 7 19:51:22 2002. The cvsup man page says:
date=[cc]yy.mm.dd.hh.mm.ss
This specifies a date that should be used to select the revi-
sions that are checked out from the CVS repository. The
client will receive the revisions that were in effect at the
specified date and time.
At present, the date format is inflexible. All 17 or 19
characters must be specified, exactly as shown. For the
years 2000 and beyond, specify the century cc.
So I wanted something around 2002.01.08.00.00.00. I added
*default date=20020108000000 to
/etc/ports-supfile, then commented ports-all
and uncommented all the x11- ports. I cvsupped
the ports, and all the X stuff was reset back to how it had been in
January. I used pkg_delete to get rid of all the 4.2 stuff,
then did a make install to put the 4.1 in. No problems,
and now X is working properly. To get the ports back to normal, I
reset ports-supfile and ran cvsup without the
-s option, which I usually use. That made it recheck
every single file, and all the old ones were brought back to current.
Upgraded mrs-hudson to 4.5-STABLE. Had to edit
/etc/rc.conf to change the log_in_vain
setting from YES to 1. Nothing else needed changing (except for what
mergemaster did, of course). I built the world and kernel on
mycroft, then installed them on mrs-hudson and ran mergemaster,
all without rebooting to single-user mode, which I like to do to be on
the safe side. A successful upgrade, done, for the first time,
entirely over the network, without logging in to the console. Very
nice.
The new Galeon port, the one using Mozilla 0.9.9, won't compile. I found the answer in the freebsd-ports mailing lists archives: deinstall freetype (version 1, though the error comes from version 2) then reinstall from a freshly cvsupp