MacOS X Journal


Mon Jul 22 16:20:30 EDT 2002

OK, in an effort to make a backup of my MacOS X drive, i login as root, hold down "Option" and drag the "MacOS_X" hard drive icon on the desktop onto a folder on an external (MacOS 9) hard drive.

The Finder complains it cannot read some of the files, and then it tells me it has 8.63 GB to copy. Trouble is, the source volume is only 5.34 GB in size. Looks like it's summing the size of all the files on the source volume, rather than the space used on the volume itself; a simple mistake, but still annoying.

More annoying is the fact that it's unobvious how to make a good backup. I can't use dump, because it doesn't understand HFS+. I cannot use cp or tar, because these will not get special files in /dev, nor do they understand how to restore the HFS+ resource forks. I expect cpio will suffer the same problems with resource forks.

Searches at Apple and Google suggest that indeed, MacOS X is an operating system that is incapable of doing a full backup and restore of itself, when installed in the recommended (HFS+) configuration, unless the end user installs additional software, which pretty much means shelling out more money for Retrospect. At least the people at Dantz should be pretty happy about this. I cannot emphasize just how pathetic this is. UNIX and MacOS "Classic" can both backup and restore themselves completely. Ironically, this makes MacOS X worse than the sum of it's parts.

According to this page, out of the box, MacOS X cannot back itself up, and one should install a perl module that includes "psync." This is what i am doing. I hope it works.


Thu Jun 27 19:04:20 EDT 2002

OK, i installed "Networking Update" 1.0 and preliminary results are that it has fixed the problem i was seeing on the PBG3 with no Ethernet network after waking up looks fixed.

Thu Jun 27 02:07:15 EDT 2002

Have i mentioned how much i hate the stock automounter? It's prone to getting stuck in I/O wait (the process has a "U" state) when changing TCP/IP environments with out a reboot (as in going from Ethernet to PPP(Modem)). When this happens, the machine must be hard booted (Command-Control-Power.)

Thu Jun 27 02:06:24 EDT 2002

After updating to 10.1.5 (from 10.1.4), much of the time, after the machine wakes up, it appears to have forgotten it had a TCP/IP network configured. If i do ifconfig en0 up, things come back OK.

Thu Jun 27 02:02:08 EDT 2002

I'm having a hell of a time getting MIT Kerberos 5 to compile. From what i can tell, the MIT source was written to compile on MacOS "Classic" and so has problems decided where to look for includes, etc... sigh.

After some poking around, part of the problem appears to be in /usr/include/TargetConditionals.h :

#define TARGET_OS_MAC 1

Apparently, the above line tells gcc to compile for MacOS 9. But i'm running MacOS X, and do not even have a MacOS 9 System Folder - this install of MacOS X has never been setup to run "Classic" applications.

When i change all the occurance so the above to:

#define TARGET_OS_MAC 0

...the compile proceeds, but it's still go some more problems...


Tue May 28 14:51:29 EDT 2002

OK, i have reinstalled the MacOS X for the third time, this time into an HFS+ partition. Performance "feels" faster, and notably, Netscape 7 (pre) now works.

I'm still trying to get MacOS X to swap on a raw device. I have a second partition, 256 MB in size. From what i can tell from Apple's Darwin docs, there's a process called dynamic_pager that talks to the kernel and allocates swap areas based on need.

Of course, there's no man page for it. (Apple really needs to clean up the docs. Supposedly, this is version 10.1.3, but it feels like 10.0.1. It's functional and stable, but there are a lot of really obvious things wrong. Most of them are not serious, but still, it's a lot of really obvious stuff i do not expect to see in a x.1.3 release.

And there's no "swapctl" program, so i can't even do "swapctl -l" to see what i'm swapping on. In fact, as far as i can tell right now, there's no way from the CLI to tell where the machine is swapping.

Last night i double-clicked a huge text-file (a .smi file that MacOS tried to open with "Text Editor") and crashed the machine. I had to do Com-Con-Power. I had a session open from another host, and it was locked up. The machine responded to pings, but that was really it.

The mount command is ignore /etc/fstab. In NeXTSTEP 3.3 i would use the NFSManager GUI. No such beastie in MacOS X. Grr.


Mon May 27 23:47:58 EDT 2002

I just saw update_prebinding (a component of the MacOS Installer program) using over 300 MB of total memory (RAM + SWAP) while finishing up the installation for the Developer tools. That can't be healthy. Maybe that's why the "Optimizing Performance" drags on for ever.

Mon May 27 22:20:34 EDT 2002

Well, apparently MacOS X's Installer will allow you to do a completely useless installation. I decided to reinstall onto a disk with two UFS partitions, which required laying out the partitions by hand. Unfortunately, the installer didn't automatically create the tiny HFS+ partition that appears to be necessary to make the disk bootable. (Probably the first or second stage boot program?)

And the OS installer just chugged right along, installing into the first UFS partition that it knew would never boot. Sigh.

After picking Demetri's (who has been running MacOS X for awhile) brain for awhile, i decided to make the primary partition HFS+ and the secondary (for swap) UFS.


Mon May 27 16:36:56 EDT 2002

I Ordered 512 MB (more) RAM today from MacGurus. This seems excessive. I'm beginning to consider switching to Darwin, X11R6 and good ol' TWM.

My main reason for using any version of MacOS at all is that i have a major dependency on Quicken, in order to keep books for my (tiny) corporation in a format my accountant can handle. There's a GNU equivalent, GNU Cash, but i really don't want to have to run GNOME, which GNU Cash requires.

Sigh. If i wasn't such a lame programmer, i'd try to write my own.

Speaking of Quicken, Quicken 2002 ("Built for MacOS X") refuses to install on the MacOS X (UFS) file system. It installed on the MacOS 9 (HFS+) disk i had handy, and then i was able to copy it over to the UFS file system.

All versions of Netscape i have tried exit after their splash screen. No error, nothing even resembling an explanation. I think gangleri doesn't have enough memory for them. I guess it's going to be Omniweb or kowtow to the evil empire (Microsquish) for IE, until i get more RAM. Sad, sad sad.

I'm already sick of seeing throbbing blue buttons.

Why isn't "Eject" called "Put Away"? Mac OS 9 had both, which isn't strictly necessary, but going with "Eject" means that the way i unmount a AppleShare volume (served from another host) means i select it and do "Eject." Guh.


Mon May 27 00:05:00 EDT 2002

So i just installed MacOS X (10.1.3) today on my PowerBook G3 (pismo).

Amazingly slow, particularly the installer. The system claims to have free RAM and lots (40%+) idle CPU. Maybe it's the IDE drive. Ugh. Maybe i should just run Darwin on it...

I think NeXTSTEP 3 was prettier. I mean, it's not like MacOS X is ugly. And i'm a freak and all because my desktop of choice is TWM. But still.

NeXTSTEP on the almost comparitively powerless 33 MHz Motorola 68040 in a older (circa 1995) NeXT Turboslab was a speed daemon compared to the 400 MHz G3 Each machine has 128 MB of RAM.

Except for its inherently-crippled IDE drive, gangleri, the PBG3 "Pismo" should be the fastest machine i have. Instead, it feels like the slowest. Sad.


$Id: MacOSX-journal.html,v 1.2 2002/09/01 10:33:43 johan Exp $