Bob's big MythTV adventure. The things that went wrong, and how I fixed them. everything before Friday 19th is "remembered". Everything before the 17th is a severe blur. One lesson I learned is that IRC is a piss-poor tech support forum. Various discussion threads intermingle and step on each other. People ask a question and it scrolls off the screen before anyone competent notices. When a veteran walks into the room he can't really scan it to see whose problem is being ignored. What we really need is some sort of interactive chat where you can create and browse "trouble tickets". Supplicants would create a trouble ticket and wait for someone to chat with them about it. Veterans would listen in on the buzz from all active discussions, or browse trouble tickets and enter into focused discussions. - mythtv@thoth.purplefrog.com November 2004 Friday 12 - Monarch Computer completed assembly of the system. I picked it up and took it home. I started out calling it Aylee (monster from www.sluggyfreelance.com, much like Bun Bun), but after a few days changed it to Alexandria to conform to my Egyptian theme. I installed the base gentoo system and then left it emerging a pile of crap while I went to dinner with friends. (The pile of crap was culled from /var/lib/portage/world on BunBun, my first MythTV box which was too slow to actually use, but served admirably as a test platform and gentoo teething box). The compile broke partway through on some packages. I wrote a script to flog it past the errors and left it compiling for the night. The IRC channels on irc.freenode.net (#mythtv-users, #gentoo) will prove to be at various times vital and useless as I continue this adventure. Weekend 13 & 14 - Mad rush to get everything working! I forgot to "emerge -u system". I can't remember how I got around the perl module that wouldn't compile. I think I just changed the USE flag that caused another package to want it and left it for dead on the roadside. I have the Hauppauge PVR250 and ATI TV Wonder VE in the case. on-board audio doesn't work. The USB mouse doesn't work. The system crashes a lot. It's DOG slow. I applied the wrong tuner=50 patch and had to diff my kernel tree against Bun Bun to see that some of the frequency multipliers were wrong. I fixed them and now IVTV works. I eventually recompiled the kernel with NForce IDE drivers enabled. Much faster. Still crashes. I abandoned the Mad Rush, disabled a lot of stuff, and started a Slow Progress strategy, fighting one devil at a time. I configured X for 1024x768. Mouse doesn't work. I do most of my work by ssh-ing in from my laptop, using alexandria's monitor only to run mythfrontend and mythsetup. I gave up on motherboard audio and pulled an ES1371 out of one of my older machines. Mon 15 - Running with DMA disabled and only the ivtv tuner. I also disabled commercial flagging because it was the last message I saw before a SEGV. No real problems. Mythtv recorded several shows. Tue 16 - I enabled DMA, expecting it to crash. I discovered it crashed in the morning, but I hadn't run the daemon in a VNC session, so I couldn't know why it went boom. I restarted the daemon and didn't have any real problems after that. I installed a Hauppauge DVB card loaned to me by work just to see if I could get it working. I installed the drivers, but had no frequency table for Charter. I used their web site to "request support", but I doubt they'll give me a frequency table. I ran into some gayness with the X server -vs- AGPGART. This would cause the X server to lock up when I tried to play video. I disabled the agpgart module and things worked better. [ A day or two later the agpgart module reappeared (for reasons I can't remember) but then it works. freaky. I'm probably still running with it enabled ] I got MythWeb working. It took me about an hour to figure out how to enable mysql support in mod_php. The crux of the problem was that I didn't realize mod_php was the issue (or even that it existed). I recompiled PHP and flogged it for a long time. #gentoo was useless. Wed 17 - finally got USB working. I had overlooked a module in the kernel that was VITAL. nvaudio still crashes the computer. I installed LVM and created an XFS partition for /home/video. I copied all the files over. I hope to do the same for /home, /usr, and /var (although with reiser instead of XFS). I yanked out the DVB card and put the bttv card back in. I hooked up Alexandria to the Dell 2300MP projector. The projector doesn't appear to be synced up with the video card very well. Some things that should be sharp pixels are blurred. I finally figured out why the ivtv module wasn't being autoloaded, despite the correct entries in /etc/modules.conf. /etc/modules.devfs had incorrect entries referencing /dev/video* which was shadowing my /etc/modules.conf lines. Deleting those lines from modules.devfs solved the problem. #gentoo was useless in resolving this problem. Thu 18 - I left for work with the cable plugged into the bttv card. It recorded a lot of static while I was at work. USB doesn't work with lilo. I also discovered it doesn't work with GRUB. I'll have to poke around in the BIOS and see if there's some sort of "usb keyboard PS/2 emulation". I started playing with bttv again in earnest. I can't seem to capture sound. When I finally downloaded sox I was able to verify that btaudio was useless (my card was documented as "not having audio encoding" no the web). A patch cord to the sound card line in enabled me to hear the sound. I downloaded many different mixers (cmix is my favorite command-line tool). Signal quality is ass when I use a second splitter to supply signal to the bttv card. I blame the really long, really old coax cable I'm using. I'll have to buy a better splitter. In the mean time I pulled the VCR out of the entertainment center and used it to convert a composite A/V signal from the PS2 to a coax test feed for the bttv card. Fri 19 - I finally re-read the MythTV howto that says "You can not use xawtv to determine if your audio is working correctly". fuck. Thursday evening was wasted. #mythtv-users was useless in resolving this problem. In my mythtv config tuner 0 is the ivtv and tuner 1 was the bttv. I have a theory that because the ivtv card referenced /dev/dsp in its mythtv config, that tied up the device and prevented tuner 1 from using /dev/dsp. I reconfigured mythtv and made tuner 0 have a blank sound source configuration. For the record: mythsetup in VNC is HELLA SLOW. The detailed subtle backgrounds prevent VNC from using any compression, and it basically transfers the ENTIRE UNCOMPRESSED FRAMEBUFFER when the menu changes. Plus the refresh strategy of the mythsetup app has some egregious stupidities (like it switches to a pristine copy of the wallpaper before the app finally exits). Others have seen my tuner=50 bug report in sourceforge and one emailed me about it. I hope my response helped him. The tuner reconfiguration solved the problem. MythTV can now record audio patched from my bttv card. Unfortunately, the backend is still a little confused (didn't change the channel to make some recordings, so I got the wrong program). I had to restart the back end. I found the 4-port cable splitter in the drop ceiling RIGHT ABOVE MY RACK. How convenient. Now both tuners are getting as much signal as my ReplayTV was before I started splitting. Weekend 20 & 21 - I experimented with manual commercial flagging. Something is going wrong with the transcoding, and I haven't investigated at all. The MythTV box locked up a couple of times this weekend. Sometimes it was just that the back end went insane and I could restart it. Other times the machine was locked up with something displaying on the monitor, sometimes without, and required a front button Reset. I missed some shows I cared about. I started fighting with the IOGear KVM switch. It doesn't play very well with the Gyration wireless keyboard. ScrollLock ScrollLock doesn't switch computers. I have to use the Numlock '-' '1' Enter, which is unreliable, either because of timing, or lost keystrokes (there are so many cables near my computers that I'm sure it's a very RF unfriendly environment). Also, my Gyration keyboard is starting to have the same key-sticking problem that plagued the first one that I returned for exchange (out of the box, at least 1/4 of the keys on that keyboard would stick down; it was heinous). Right now it's just the N. Mon 22 - I disabled the bttv tuner in the hopes that it will put an end to the lock-ups. I filled out a tech-support web form at gyration about the sticky N key, and another one about the ScrollLock not working with the IOGear. There were several instances of this in my mythtv backend log file: DB Error (delta position map insert): Query was: INSERT INTO recordedmarkup (chanid, starttime, mark, type, offset) VALUES ( '1054', '20041122123000', 3577, 6, "980490278"); Driver error was [2/1034]: QMYSQL3: Unable to execute query Database error was: Incorrect key file for table: 'recordedmarkup'. Try to repair it google helped me find the necessary mysql command: "repair table recordedmarkup;" once I picked the correct search terms. IRC was useless. I hacked up a hyper-primitive perl script that would allow my brother to access the shows I have recorded. Using a USB keyboard borrowed from work I determined that the gyration was at fault. Alexandria didn't crash all day and night. Tue 23 - I was watching some recorded programs all evening. After midnight I discovered that the mythbackend had ceased to record sometime after 7:30. Port 6544 accurately reported what was planned to record, but it had never really recorded any of them. QMYSQLResult::data: column 16 out of range QMYSQLResult::data: column 17 out of range QMYSQLResult::data: column 18 out of range I restarted the backend and it immediately started recording TV. I missed the entire prime-time schedule. The bttv card was in the case, but its module isn't even loaded, and it's not referenced in the mythtv config. Wed 24 - I ran a REPAIR TABLE on each table in the mythconverg database. No problems were evident. Fri 26 - I installed MythMusic. Originally it scanned my tunes and built a tree using ID3 tags. This was a horrible mistake because most of my MP3s predate the invention of ID3 tags (I encoded my collection back in 1999, and haven't added tags because I had no use for them). I hunted through the source and found an option that could be adjusted in the GUI setup to display things by directory structure instead of ID3 tags (and the help string for that field even documented it). Problem solved. I eventually found the keybindings to cycle through the various visualizations available. Mon 29 - I emerged MythDVD. It displayed the FBI warning screen and two vendor splash screens of my Zim DVD and then returned to the MythVideo menu. I tried the FLCL volume 2 and it didn't display ANYTHING before returning to the menu. I tried recompiling mplayer with modified USE flags. I gave up and switched to ogle. It interprets my DVD menus just fine. It doesn't seem to have the ability to play video backwards, though. December 2004 Sunday 5 - I bought parts from Fry's to create a serial cable to connect to my projector. I soldered the DB9 to a PS/2 cable and got it working from my laptop. Unfortunately, I don't have it working on alexandria. If I knew what was different between the two I could fix it. The laptop and alexandria can communicate via null modem just fine. If I can ever get it working, I'll be able to put away the projector remote and turn it on and off via software triggered from my wireless keyboard. Sun 12 - I figured out that the extraneous connections I designed into the serial cable were interfering with signalling. I determined which connections to cut, and got a working cable. I can now power on my projector using a program triggered by a window-manager key binding. Wed 15 - I installed a Maxtor 300G hard drive onto the primary IDE channel as a slave to my Western Digital. This would turn out to be a big mistake. While I was at it, I removed the bttv card. I'm already recording more TV than I can watch, and I really don't need the heat. Actually, my basement is cold, but I'll instead use the air conditioning to turn electricity into heat. I moved all the logical volumes onto the Maxtor, crashing along the way. Thu 16 - I rebooted alexandria after repartitioning the WD. It didn't come back up because /var (a reiserfs) was corrupted. I originally misread the errors and assumed /home/video was corrupted. In retrospect I'm going to blame it on the WD+Maxtor interaction. Fri 17 - Compiling mythfrontend on a laptop with 256M RAM and no swap fails. It seems some of the compiles REQUIRE more RAM. Sat 18 - The dramatic increase in the number of videos that have glitches and other drop-out induced problems is beginning to alarm me. The worst part is that I can't seek through the videos to skip past commercials. I've correlated it with the following messages from the kernel (dmesg, /var/log/messages) ivtv: ENC IRQ OVERFLOW: #703 Stealing a Buffer, 256 currently allocated google searches mention contention on the PCI bus, and hard drive performance. I moved the Maxtor onto the secondary IDE interface and disconnected the DVD drive (for testing). Sun 19 - I created a striped LVM and started moving movies to it. This exacerbated the ENC IRQ OVERFLOW problem. The IRC channel gave me pointers to the list messages about Maxtor and WD drives fighting. http://www.ussg.iu.edu/hypermail/linux/kernel/9908.3/0741.html I moved all the movies onto a Reiser LVM. Then I moved all logical volumes to the Western Digital and vgreduce-d /dev/hdc4 (formerly /dev/hdb4). I created a new volume group with the Maxtor to store DVDs, but with the DVD in alexandria disconnected, I had to use the DVD drive on moeris (the laptop). It was getting read errors really early in each of the "files" on the DVD, so the images were useless. Alexandria started misbehaving. Restarting mythbackend didn't work. I eventually traced it to corruption in mysql databases. Using REPAIR TABLE I eventually got things working again. I saw some all-zero blocks inside the database files. I don't know if that's a symptom of I/O corruption, or if mysql naturally creates files with holes. Now I'm worried that this little Maxtor incident has corrupted system files. Not only did I do a pvmove onto and off of the Maxtor, I did an "emerge -uD world" while the Maxtor held the data. The mythtv decoder software has flaws that prevent mythcommflag from properly analyzing MPEGs that have dropouts and glitches (back to the ENC IRQ OVERFLOW issue). I ran it many times manually with suggestions from IRC folk. Mon 20 - I cranked back the DMA level on the Maxtor to udma2. Somebody on the irc channel pointed out that it was a bad idea for my kernel to be optimized for my CPU. If it was too efficient, it witness some bad electrical design on the WD hardware. http://lkml.org/lkml/2000/1/30/17 I recompiled as i386. I also copied all my Zim DVDs to the Maxtor. I was unable to read the movie data (kernel IO error) unless I first launched ogle. I think ogle had the logic necessary to prime the drive with the encryption keys. I wish I had a command-line version of that. Throughout this process I didn't get a single ivtv buffer overflow. Tue 21 - There were some errors in the log when I checked this morning: 00:33:43 kernel: __alloc_pages: 4-order allocation failed (gfp=0x20/0) 00:33:51 kernel: __alloc_pages: 5-order allocation failed (gfp=0x20/0) I guess I ran out of RAM and swap from having too many programs running. January 2005 Wed 5 - I was clobbering alexandria with downloads and triggered some more ENC BUFFER OVERFLOW errors. Thu 6 - I kludged a script that I run with the mythfilldatabase. Its job is to rewrite the program table entries for CNNHN to enable me to record morning news the way I want to. Tue 11 - My CNNHN kludge script appears to be working well. I've been able to watch "A Morning Bob Show" (my fake CNN show) almost every morning. I think that if you delete a recording that is in-progress, mythtv will record it later. The recording only counts as "previous" if it completes. Wed 12 - After a couple of days of playing with DVD authoring, I actually burned a disc containing two 1-hour TV shows, each in its own chapter. It wouldn't play in one of the DVD players at work, but played fine in another. Maybe I'll figure out menus later. Sun 16 - I have the Gyration GT wireless suite: a mouse and a keyboard. My basement is a PISS-POOR RF environment for these toys. The keyboard loses keystrokes often enough that you can't really touch-type even a single word with any confidence. As a remote control for mythtv, it is usable, but as a keyboard, it's utterly useless. February 2005 Thu 3 - I installed a DVI adapter for my ASUS motherboard. I bought it from newegg a long time ago and it didn't match their photographs. It does, however, match the ASUS web site photos. I still need to buy the SVideo/Composite bracket (model AV/S). I also need to TEST this DVI adapter. Sun 6 - I installed a 512M stick of PNY RAM. It didn't work. The machine crashed. The X server comes up with mostly-white lines on it. Wed 9 - I upgraded the BIOS to rev 1004 by creating a bootable DOS CD with the ASUS tool and data. This did not make the RAM work. Wed 16 - I tested the DVI board for the ASUS A7N8X-VM/400. It has almost no logic of its own. It just provides a DVI port for the motherboard nVidia graphics. If the machine is booted with a DVI cable inserted it mirrors the video until you start the X server, after which the VGA port is useless. I discovered many things were broken after a reboot: My Gyration USB keyboard was non-functional. A day or two ago I did an "emerge --depclean". It turns out that removing "coldplug" was a mistake. Other things broken by the depclean include xterm (was missing libtermcap) and mplayer (was missing libdv). I will have to do a LOT of investigation and experimentation to figure out if it is even POSSIBLE to have both the VGA and DVI port active at the same time. This isn't critical to my plan. Having the Svideo port work at the same time as one of the computer outputs would be enough. Fri 18 - I wrote a command-line script to ldd every executable in my path and check for missing dynamic libraries. The following packages had been broken by upgrades: mplayer, mjpegtools, transcode, xawtv (I have since discovered that revdep-rebuild is the gentoo way to resolve dynamic linking roadkill. You have to "emerge gentoolkit" to get this vital tool). Mar 2005 - Sat 19 - Over the past week I've been running real low on disk space. In fact, in one instance I ran OUT. In the course of upgrading my gentoo packages I restarted mythbackend. *BAM* 60G free. The backend had some descriptors open, keeping files alive after they had been unlinked from the filesystem. I wonder if 0.17 fixes that problem. I'm not willing to upgrade just yet. June 2005 - Sat 18 - It is now non-trivial for me to remain at version 0.16. The ebuilds have been retired from portage and I had to download them from the CVS attic in order to "emerge -u world". Today I upgraded my system to version 0.18 of mythtv. The process appears to have been painless. Whatever database changes were required happened magically. My recordings were preserved. New recordings appear to have happened properly. The Time Stretch feature ('A' key) is fun and useful with CNNHN. I wrote another database kludge script for The Daily Show many weeks ago. It is a pain in the butt because Comedy Central re-airs it several times a day. I eventually figured out which ones were repeats and which were new and wrote a script which generates episode IDs, ensuring that I don't record duplicates. I have had to tweak it once already when Comedy Central changed the schedule. Wed 29 - I noticed that the :6544 web page wasn't predicting the morning recordings of CNNHN. After talking with the folks on the IRC channel, I restarted the back end. That resolved the insanity. Thu 30 - The insanity is back. I think it's related to the fact that mythtv re-downloads "tomorrow's" schedule data to catch last-minute changes. running a mythfilldatabase (which accomplishes nothing) sends the signal to the back end to trigger a reschedule. Aug 2005 Thu 11 - The folks on IRC clued me in to the "mythbackend --resched" command, which I now use to tell the backend to recompute its schedule after my CNNHN and Daily Show kludges rewrite the database. It's faster and better than running a second mythfilldatabase. Thu 25 - I tried to upgrade to 2.6.12, and consequently had to upgrade to ivtv-0.3.3. The result was video captured at too low a rate, resulting in a playback that is high-speed. http://sourceforge.net/tracker/index.php?func=detail&aid=1187550&group_id=73219&atid=537126 I reverted to 2.4. I'll try other versions of ivtv when I have more time. Sep 2005 Wed 28 - I emerged ivtv-0.3.8 for my 2.6.12 modules and it worked. Thu 29 - While watching a recorded show in the morning, I encountered severe visual flaws. Every once in a while the video would slow down, and the audio would become choppy, leading me to believe it was capturing too many frames. I configured lilo to boot into 2.4 and scheduled a reboot. Fri 30 - Even in kernel 2.4 I'm getting glitches. I switched out the new firmware for the old and that solved the problem. Oct 2005 Sun 2 - I rebooted into kernel 2.6 still configured with my old ivtv firmware. It looks good. Now I just need to get the PCHDTV card working. Mon 3 - I ugpraded apache. The conf files have been moved. I had to re-emerge mod_php and make sure -D PHP was in gentoo's apache start-up config. Thu 6 - I'm still stumbling with this 2.6 upgrade. I created a couple of versions that didn't work, and it appears it is because ivtv was no longer being autoloaded by the kernel when mythbackend asked for it. I have arranged for the init process to load the module instead. I still don't have the PCHDTV working. Sun 23 - After banging my head against the wall for a while, I finally found the proper driver for my PCHDTV 3000 card. Unfortunately, the cx88xx_dvb module requires the tveeprom module, and the kernel tveeprom conflicts with the ivtv modules. I had to remove ivtv from the /lib/modules directory. Then I was able to use the instructions at http://www.avsforum.com/avs-vb/archive/index.php/t-530806.html to tune some channels and capture some video using the command line. /usr/share/dvb/zap/atsc/make_atsc_chanconf.pl azap -r WAGA dvbtraffic The channel list it came up with was kind of short, so I went to antennaweb.org to add some channels that will require my directional antenna. http://www.antennaweb.org/aw/Address.aspx I rebooted to a stable kernel so ivtv would work again. I'll have to wait till the module conflicts get sorted out before I try integrating HDTV into mythtv. Thu 27 - Over the past few days I've been noticing some serious visual glitches in recordings where single frames have displaced chunks in the lower half. This glitch is reproducible in mplayer. My first instinct is to blame ivtv 0.4.0. Dec 2005 - Dec 7 - I've been running with ivtv 0.3.8 for at least two weeks now. No more visual glitches. I switched my sound output to ALSA. I changed the sound output from /dev/dsp to ALSA:default and the mixer from /dev/mixer to "default". This didn't magically allow me to play a sound at the same time mythtv is playing a show. May 2006 - Mon 8 - I added a seagate 300G drive to the mythtv and have been expanding my LVMs for a while. I've been getting more and more audio buffer underruns lately. It turns out I hadn't configured DMA for that drive. hdparm -d 1 /dev/hdb solved the problem. I updated /etc/conf.d/hdparm to make sure that it would activate DMA after every reboot. Fri 26 - I finally figured out what was wrong with the TV out on my old laptop. I had configured Xorg for 680x480 display resolution and there was no modeline for that. I configured it for 640x480 and it worked. I now can use that laptop as a mythfrontend to drive the Sony television when I want to play video games on the projector. June 2006 - Sat 3 - I upgraded to PHP 5. It broke mythweb (probably some obscure configuration option for apache). Sun 4 - I ugpraded to MythTV 0.19. I get colored ghosts in the preview window on the program selection screen. There appears to already be a bug in mythtv's tracker about that. I fixed mythweb/PHP. All my bookmarks for the mythweb interface are now broken by the new layout. I have to upgrade my frontends from 0.18 to 0.19 now. Thu 22 - A couple of days ago alexandria balked during a reboot, unable to fsck its /var. Powering down for a minute made the problem go away. I worried that the hard drives were overheating. I bought a drive cooler that mounts to the underside. I started doing lots of pvmoves to prepare for a reconfigure of LVM onto RAID5. I got a couple of lock-ups, but nothing I couldn't work past. Today I came home from work to see that a pvmove had locked up the machine. This time the reboot didn't work. /dev/hdc (a Maxtor) was quite inaccessible (no partition table). Turning off DMA made it work (ide=nodma kernel argument). I am pvmoving all data off /dev/hdc (v e r y s l o w l y). I hope it is still under warranty so I can RMA it. In the mean time I'll build the RAIDs using the 200G WD, the 300G seagate, and a 200G (WD?) I was using as my laptop backups hard drive. Thu 29 - The pvmoves are complete. I'm running with a few mirrors and RAID5s spanning from 3-4 drives. I lost some recordings as blocks got corrupted during the pvmove, but nothing to cry about. I now believe that if a hard drive loss would even be inconvenient, you should use RAID, even slow software RAID5. I have cooling fans strapped to the bottom of two of the hard drives, and the other two hard drives are in a position to get incidental airflow from those fans. August 2006 - Fri 11 - The mythbackend froze up. There is a possibility it was triggered by my upgrade to gcc-3.4 (lots of recompiles). Cctober 2006 - Sat 14 - Upgrading to kernel 2.6.16 and ivtv-0.6.3 resulted in mythtv being unable to tune to live TV (and I assume unable to record shows). Removing the tuner via mythtv-setup and re-adding it and re-associating it with the cable line-up solved the problem. Sat 28 - I replaced a 160G hard drive with a 300G hard drive and hot-added the partitions to the RAIDs. Reconstruction proceeded in the background. Mon 30 - The mythfrontends were unable to communicate with the backend. I restarted mythbackend and checked the system logs. Something is wrong with a reiserfs. Tue 31 - I did an fsck.reiserfs during an idle period and the tree on /home/video was corrupt. ########### reiserfsck --check started at Tue Oct 31 16:44:02 2006 ########### Replaying journal.. Reiserfs journal '/dev/vg00/lv_video' in blocks [18..8211]: 0 transactions replayed Checking internal tree../ 3 (of 124)/ 41 (of 168)/ 1 (of 85)bad_leaf: block 34677, items 1 and 2: The wrong order of items: [2 281 0x0 SD (0)], [2 279 0x0 SD (0)] bad_path: The left delimiting key [2 278 0x1bb001 IND (1)] of the node (34677) must be equal to the first element's key [/123 (of 124)/121 (of 135)/ 1 (of 167)bad_leaf: block 30622729, items 1 and 2: The wrong order of items: [2 280 0x0 SD (0)], [2 279 0x1 IND (1)] bad_path: The left delimiting key [2 279 0xd1fbc6f6001 IND (1)] of the node (30622729) must be equal to the first element's key [2 278 0x1bb001 IND (1)] within the node. finished Comparing bitmaps..finished Fatal corruptions were found, Semantic pass skipped 4 found corruptions can be fixed only when running with --rebuild-tree ########### reiserfsck finished at Tue Oct 31 17:04:40 2006 ########### I deactivated mythbackend and performed an fsck.reiserfs --rebuild-tree /dev/vg00/lv_video It took many many many hours. It appeared to be stalled in pass 2 for the majority of that time, although strace revealed it was doing a lot of disk IO. Nov 2006 Wed 1 - The fsck.reiserfs was still running when I woke up. I created another 100G filesystem to take the place of /home/video while it was being rebuilt and mounted it. I discovered the audio was being weird. I used mythtv-setup to remove the HD card and removed the kernel modules, then reactivated mythbackend. That solved the problem. I've read things about the 2.6.16 kernel and ivtv -vs- the PCHDTV so I'll have to re-read it and do the fixes. The rebuild-tree finished after over 12 hours of disk IO. Dec 2006 Mon 2 - In the process of recompiling my system with gcc-4.1.1 I have driven my ivtv drivers insane. When I ask it to tune to channel 27 (CNNHN) I get Lifetime. Channel 73 is not SciFi. Channel 52 is not Cartoon Network. Recompiling the kernel didn't eliminate the problem. May 2007 Mon 28 - I have been having trouble with missing packets in my recorded shows. Sometimes its just a glitch at the beginning. Sometimes it is mangled in the middle, driving the mythtv player insane and sending it into fast-forward mode. I upgraded my kernel to 2.6.20-gentoo-r8, and my ivtv to 0.10.1-r1. The ivtv ebuild had many complaints about my first kernel configuration, but it spelled them out pretty clearly and I fixed them. It appears to work. We will see how mangled future recordings are. Nov 2010 I've been running with 10-40G free for months. I need to either give up on some shows, or catch up. Mon 8 - I finished my upgrade to MythTV 0.22. I had to do a lot of work to make the database schema upgrade work. The 1170 schema update seemed to be quite non-deterministic (sometimes it would abort, but eventually it succeeded), possibly because it constructed an IN clause using string concatenation. I also ran afoul of the mysql character encoding issue. qt-webkit couldn't actually compile on alexandria because it required too much RAM (severe thrashing). I built a binary package on my laptop and copied it over to alexandria. Now I just need to update all the OTHER software on that host. Plus I have to update mythtv on my other hosts. Since I have neglected updates on them this could take a few days.