Older blog entries for spot (starting at number 112)

12 Apr 2003 (updated 13 Apr 2003 at 06:28 UTC) »


cvs glibc rpm vs spot:

cvs glibc rpm: 18
spot: 1

This is driving me nuts. When I build it from source, by hand, on the commandline, it works. When I do it from within RPM, it dies randomly, in random places. :/

i only needed to score once to win. :) I've now got working, recent glibc rpms.


Fry died. Fry was my build box. It used to be an AXi.

I put Fry's hard drives in my U60 (that had no working drives), and Fry lives again. The only downside is that the new Fry has 1/4th of the ram that old Fry did (anyone know where i can get cheap ram for a U60). The upside is that it has two CPUs instead of one (2 X UltraSPARC-II 360MHz).

Also, none of the nextgen Aurora work was being done on Fry, it was all on my SunBlade 100.

No real Aurora code work to mention until I get into Chicago in a week. All the important machines are in piles.


Jakub is my hero. We've got a working elfutils now, and he pointed out something fairly obvious about why perl exploded.

We need to come up with a name for the Aurora code-in-progress branch, e.g. rawhide, but not rawhide, so that I can start synching that tree out to the world at some point in the future.

bugzilla.auroralinux.org is up and running, I'm going to announce it to the users shortly, and put it on the website. it should make my life a little more organized. I love the increase in users, but it makes the mailing list a difficult place to track bugs.

Real life:

Moving to Chicago on Feb 15th.


I spent a good chunk of the afternoon learning how elfutils worked, added some code to make it build for sparc, then realized, it didn't get me past the error that was stopping the compile. I asked Jakub, and he pointed out the obvious answer. I'm eternally grateful for his patience on stupid questions. :)

It only fails four of the tests now. Its a start. :/


About 3 months ago, I realized that I was drinking about 6 or 7 cans of mt.dew a day (the fact that they are free in the office definitely wasn't helping matters). I came to the conclusion that this was really unhealthy, and that I should do something about it, so I cut caffeine cold turkey.

Needless to say, that was unpleasant. I had vicious headaches for about a week straight, but then, they eased off. About a month later, I kicked soda altogether (the only exception, a sprite at one of the hockey games). Water and gatorade have become staples for me, since I'm constantly thirsty (theres a 32oz water mug on my desk that gets filled once a day), but the random headaches that I was getting while I was on the caffeine have gone away, along with some of my moodiness. I've definitely been more focused without it, although, I haven't really been able to pull the sort of late nights that I did a few years ago. I don't really miss it, but I didn't drink coffee, so that helps, I guess.


I was promoted last week to an outside sales engineering position. I'll be based out of Chicago, and covering the midwest territory. This means a relocation to Chicago sometime in February. Most of the fine details won't be settled until tomorrow.


Work on the 8.* build is not going terribly well, mostly due to the fact that I can't get glibc greater than 3.2 to build/glibc 3.2 won't build for 64bit because I can't get gcc 3.2 to build, because it needs glibc built for 64bit. Do you see my problem? ;) I could just be braindamaged on this one, Jakub says it works for him, he was going to build it overnight.

Dave Miller is doing the NPTL work for sparc64, which should be a nice feature addition. I think I owe him a beer or twelve.

2.4 seems to be unhappy on the sun4 arch. If interested, see my post here. Yes, I suppose that does make me a masochist.

The Vera fonts that were donated to GNOME by Bitstream are very nice, imho.


mini-Aurora update:

starting the initial work on the 8.* tree. forgot how much pain it really was. :)

Well, damn. Advokitty is fast for once. I haven't posted here in a while, been busy. Here's the current musings:

Over the holiday break, I got Red Hat Linux 8.0 running on my ProGear touchpad tablet PC. Only thing not working is the PCMCIA. When I get bored again, I'm going to rebuild the kernel to see if I can fix it.


I put my foot down and said that there will be a new release in January, because if I didn't, it would keep being tuned and tweaked indefinitely.

Fixed the O(1)+low latency implementation for sparc64 (#1) so that it worked again. Decided to include it in the next release.

SS2 at work is really on its last legs. Its doing odd, unreproducable bad things. Going to take it out back and shoot it...err... return it to the donor as soon as I can replace it. Chris Hedemark donated a LOT of older Sun hardware to Aurora, so I need to go through that pile and perform triage on it. Best item: a WORKING (#2) Sun 4/300. No, I'm not really planning on supporting sun4 in Aurora. That would require far more than its worth. I just like old hardware, sue me.

There be dragons in the sparc64 VM. Thats my guess. It could be a bug in the O(1) implementation, but doing a mke2fs call in the GUI mode of anaconda (not TUI mode, that works fine) hard locks a sparc64. Hard lock as in, can't drop to prom. How do we resolve this? We run it through strace. It works then. Say it with me, Voo Doo.

Large disks partitioning/formatting in GUI mode (again not TUI) also hard locks the system. Even with mke2fs being run through strace. I'm going to generate a build with parted also run through strace and see if that resolves the issue.

Yes, I do realize thats incredibly hackish and bad.

Onboard network adapters aren't being properly autodetected on the U10, but the driver works fine, and the proper entries are in the kudzu sbus probing code. Same goes for the SCSI controller on the SS2. And some of the sym53c8xx controllers (SunBlade 2000). I can't find a good reason why.

Fixed USB implementations in both GUI and TUI. TUI didn't work at all in the last build on USB driven machines. Works almost perfectly now. Mouse isn't being autodetected, still can't pin down exactly why, but it works when you choose Generic 3 Button (USB).

The "parted" bug in 0.42 is very dead, the stock parted code for sparc is, well, bad. No offense to Ben C, but ick. It was misreading a huge subsect of disks. I don't claim to have made the code any cleaner, but at least the Aurora parted works on every disk I can throw at it now.

Sound detection is a little weird on some boxes. *shrug* sndconfig tried to load the sun4c audio driver on the SunBlade 2000. Shockingly enough, it didn't work. Although, everything else worked like a charm. It ran like a bat out of hell. :) I just wish it was my machine to keep.

Added cipe to the kernel. Didn't think it would work, but it seems to have built ok. It works on alpha, so it should work on sparc. I'm going to let Ingo test that.

The tree is too large to install off a single CD now. *shrug*

The GUI installer does bad things when dealing with Ingo's SS20 video card, because it has more memory than the standard, and I don't really have a good way to probe for it. Maybe if I had it in one of my machines, and another month... :/

Sun4d still doesn't work.

I really need a second apartment just for all this sparc stuff. I'm not complaining, just stating the obvious.

Looking at the small amount of sparc/sparc64 kernel code that I grok, reminds me how much I admire Pete & Dave (and Rob) for hacking in there. I'm not really a programmer, I just play one on TV. Those guys are programmers.

No one is still reading this. Apologies for the long entry.


And finally, release. Getting the multi-boot code to work (so I could have one ISO that boots both sparc32 and sparc64) was a little uglier than I'd hoped, but it works reliably.

I'm going to take the portable SCSI cdrom drive home and test on some other machine types, but all the hardware in the pile at work, well, works fine.

Thats a U2 (smp), a U10 (up), an SS20 (smp) and... the SS2 (up). GUI install on the SS2 even. It doesn't know what driver to use for its SCSI card (esp), but its easy to select from the list.

Sun4d doesn't work at all. But thats the kernel's fault.


At this point, I'm left with the mouseconfig failure (I need to debug kudzu to find out whats wrong with it), the KDE audio failure (not showstopping for a beta), and KDE being slower than Debian Stable's release cycle.

Video works for my ss20 (GX), and my Ultra 2 (Creator 3d). Installing without /boot partitions works. Brian Ferris's keyboard "us" bug in text mode is fixed. All the devices are being probed and detected properly. All thats left to test is the merging of sparc32 & sparc64 on a single iso set. And I feel really good about that.

8bit GUI installs are ugly (not Aurora's fault, 8bit is inherently ugly), but they work.

Its good stuff. I'm really pleased.

26 Aug 2002 (updated 26 Aug 2002 at 23:47 UTC) »


Really fixed sparc32 (and probably sparc64) hardware probing this morning. Looks like it sees the videocard properly now too. Added some heuristics to tell anaconda to always try 24 bit video, since I'm stumbling all over sparcs that can't handle 16bit. It might be slow as hell, but at least it won't crash. (and 8bit is ugly, but possible)

It doesn't even pause in stage 1 now, except to tell me that its autoloading the scsi driver. Its so beautiful. :)

Aurora Bug List:

- mouseconfig seems to be failing noisily (but non-fatally) during post

- KDE is trying to use /dev/dsp, failing. Probably can be fixed with a symlink to /dev/audio.

- parted is complaining that the H/S/C counts on the sparc32 disks (older disks?) don't match what the disk label says... parted is wrong, needs patching.

- parted/silo seems to fail on machines without a /boot partition during install, although, the silo.conf generated seems valid

- KDE/QT is being poorly optimized by gcc 2.96, causing KDE to have terrible performance (5 min app load vs 30 sec for GNOME). This is not an issue with gcc 3.1.

- GUI installs always try 16 bit, and fail for majority of Sun video hw, added heuristics to anaconda code to try to make it smarter with this regard, for sparcs. Partially complete.

103 older entries...

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!