< | >

Hacker's Diary

A rough account of what I did with Emacs recently.

December 31
Minor tweak to BBDB to fix a buglet I introduced while attempting to fix the VM problem. The VM problem still exists, alas; I thought I'd gotten rid of it.

Booted Gonzo with a 2.2.18 kernel before I left for New Year partying. It was still running sans filesystem bug when I got home. HMMM.

December 30
While looking up the Linux Softmodem stuff, I stumbled across a new driver for the Lucent winmodem I have on board, which has compilable kernel support and a binary driver. This means you can upgrade your kernel and still have a working driver. Well, in theory, anyway. When I finally persuaded it to link and all that, it caused a kernel panic. Bleah.

December 29
Yay! Got my IrDA port working on the Windows side of the house. Windows had somehow managed to install 4 IrDA devices, all of which fought with each other. Clearing them all out and redetecting fixed the problem. However, the Ericsson software for my phone wants me to have Outlook installed, not Outlook Express, or at least something with mapi32.dll. And the modem via cellphone doesn't appear to work at all.

December 28
A little more tweaking of X (thanks to reading the log files, this time!) and yay, 1024x768. The LCD panel is 65Hz, not 60, but XFree86 4.0 seems to be smart enough to figure out the refresh rates itself. Keen! Heck, even the resolution-switch keys are working, which I've not had on any machine since I first started using Linux. And I can switch from X to VT to X again, which is something that I was missing since I installed XFree86 4.0. Hoorah! Next thing to clean up is the syslog config, I think...

Klortho seems to be okay with the new kernel, but Gonzo's still doing the failed-ln thing:
brk(0)                                  = 0x804d96c
brk(0x804d994)                          = 0x804d994
brk(0x804e000)                          = 0x804e000
stat64("bar", 0xbffff9a8)               = -1 ENOENT (No such file or directory)
lstat64("foo", {st_mode=S_IFREG|0644, st_size=0, ...}) = 0
lstat64("bar", 0xbffff938)              = -1 ENOENT (No such file or directory)
lstat64("bar", 0xbffff938)              = -1 ENOENT (No such file or directory)
link("foo", "bar")                      = -1 ENOENT (No such file or directory)
write(2, "ln: ", 4ln: )                     = 4
write(2, "create hard link `bar\' to `foo\'", 31create hard link `bar' to `foo') = 31

I really, really, REALLY hope this isn't a motherboard issue. Then again, maybe I should get myself a new motherboard for Christmas...

December 27
Reading documentation helps, but you have to read a heck of a lot of it. Finally I have a mostly-working XFree86 4.0 setup, modulo the fact that it's coming up in 800x600 instead of 1024x768. The machine's booting happily into a 2.4.0test12 kernel on a non-framebuffer console (I don't need that damned penguin!) with only minor problems (gpm has to be killed before running X, although I could probably wire up gpm in Repeater mode to get around this, and it may have been down to having the mouse device misconfigured anyway) and seems to be stable enough, although I'll leave my ln failure test running to see if the filesystem is okay. For some reason, Alan Cox is turning out patches for 2.4.0test13, which doesn't seem to exist - I'm obviously still missing something in this whole kernel building lark.

In the middle of all this I discovered that Gonzo had some slight borkeness in the library versions which may have been causing the problems with the disk. I'll check it out tomorrow, I guess.

December 26
Started experimenting with the framebuffer stuff on Klortho to see if I could figure out why it was breaking. It appears that the vesafb modes lock themselves to 72Hz, while the LCD is at 60Hz. Why this boots and then fails is beyond me. I ditched vesafb and started playing with aty128fb instead, and was halfway through getting it to detect my Rage Mobility Yadda Yadda when I thought to grep the patchfiles I had lying around in the kernel directory. Bingo. Someone had already done the coding, and it's in atyfb. Doh. So, I started on getting that up and running and had a frustrating series of reboots and assorted lockups. Somewhere along the line I decided to clean up my half-billion XF86Config files and managed to nuke the working one. DOH!

In between kernel builds and reboots I did a little work on automatically parsing the email for Micromail site updates.

December 25
Much Christmassing. You know you're going to the pub too often when they give you free drinks...

Not entirely without geekery, mind you. My parents' computer required a gentle LARTing to stop it from crashing on internet stuff (kids, upgrade your copy of Go!Zilla if you're getting bluescreens from IE since you installed it...) and I played a bit more Midtown Madness 2 for light amusement and to show my folks approximately where I'd been on my trip to San Francisco.

December 22
Well, taking out the optical drives has stopped the drive from falling into "click... click... click..." mode, so that's a start. The problem may be simply a lack of power - I guess I should check the power consumption or something. The ln thing is still happening, though. It kicks in reliably after about 30 minutes of uptime, but only - I think - when DHCP is active. I need to investigate that further. In the meantime, I notice there's currently a filesystem corruption bug in pre12 under investigation. HMMMMM.

December 21
Uh-oh. Hard drive started clicking last night, so I shut down. This morning, Gonzo can only find one of the four IDE devices. Time for a case-cracking, I guess.

...which revealed nothing amiss, although when I disconnected the DVD and CDROM drives, the machine booted happily. Hmmmm. I've left a script running to figure out how long after uptime the write failure starts happening, too.

December 20
Vroom! Played Midtown Madness 2 for a while, completing all the SF blitzes & circuit races which gives me, among other things, the Audi TT which seems to be required to win some races. A shame; I was trying to complete as much as I could in the '67 Mustang Fastback, 'cos it's my favourite car in the game.

Did a little more investigation into Gonzo's disk problems. It appears that after some random amount of time, I can no longer stat files on the big disk. Except that's not quite it either:

Waider (gonzo.waider.ie) $ > touch foo
Waider (gonzo.waider.ie) $ > ls -l foo
-rw-rw-r--    1 waider   waider          0 Dec 20 19:18 foo
Waider (gonzo.waider.ie) $ > echo "bar" " foo
Waider (gonzo.waider.ie) $ > cat foo
Waider (gonzo.waider.ie) $ > ln foo bar
ln: create hard link `bar' to `foo': No such file or directory
Waider (gonzo.waider.ie) $ > 
This, of course, breaks DHCP, which causes me grief. And meantime, X on Klortho drops every so often - I think it's something to do with VMware - so I'm currently thinking I shouldn't have done that piecewise upgrade.

December 19
Spent about two hours in the house, recovering from the Stepstone IT department Christmas party and getting ready for JC's "I'm quitting my job" drinks. About the only computing thing I did was to check email. Wherein I discovered that a patch I submitted to the mp3-info project had been accepted. Yay!

December 18
Not only has the drive not died, it's stopped throwing errors, too. Maybe it was cold, or something. I've built a pre12 kernel, but I've not yet installed it.

Checked in a fix for BBDB/VM once SourceForge were done with their upgrade. It appears that the windowing bug is, in fact, an emacs bug curable by doing (sit-for 0) at the appropriate point.

Reinstalled Win98 from scratch, this time under VMware. This appears to have cured all my net-related problems. Momentary bogosity with phantom comm ports and crashing IE5, but I seem to have fixed that. Installed Netscape 3.04, and was amused that this site isn't totally broken under a browser that old - obviously the stylesheet stuff doesn't work, but the rest is pretty kosher. Now setting it to install IE5.5 so I can see if the fix AjD gave me works.

Gonzo's kernel build failure was solved by nuking egcs and installing gcc-2.96 instead, btw. Which showed up a bunch of non-ANSI errors in the kernel source (shame!) where people had incorrectly used ##.

December 17
Well, the drive hasn't died, at least. And a kernel upgrade, since I'm still somewhere in the 2.4.0pre series, may well sort it out. Of course, my compiler setup on Gonzo is currently complaining that it can't find stdarg.h, and Klortho's having problems with wish. Bah.

Meanwhile, back at the ranch, er, hackery: generalised the script I use to update the site, which does stuff like cleaning up crap I don't want to upload, and refreshing the BBDB mirror, so that if I move it to another machine I'll just need to change a few lines at the top to point at the right executables. Also, the DNA Lounge layout changed slightly so I had to update snorq to compensate. Ideally, snorq would have some ideas about how to do this itself...

Upgraded my copy of VM again (Kyle's certainly in "active development" mode lately!). Made a change to BBDB to fix an odd bug I encountered related to window configurations since Kyle had already modified the VM code for this. Alas, the problem still exists.

AjD gave me some advice on stylesheets that might fix the IE 5.5 problem whereby all my text except the links is invisible. Thank you Microsoft for that.

December 16
Back from a two-week stay in San Francisco. My shiny new 30GB drive is throwing "disk busy" errors at me. This does not make me happy. I think maybe I'll worry about it tomorrow, though. Other things to worry about resulting from my bizarre upgrade can also wait 'til then.

previous month | current month | next month

'tis the season, and all that.