Yeah, no log files… I guess sleep is outside the standard Unix framework — no surprise there. (I’m still trying to find out Apple’s “official” position on how sleep functions and what happens to periodic system housekeeping as a result.

Thinking out loud –quarter-backed idea– maybe there’s something you can do to create your own log files based on some “interesting” information about the system state. Seems to me I recall something about system status and/or control in the developer tools. It might help to dump such status to a file, say, every second. Alternatively, send it to another system via LAN or even serial port at short intervals. Might show something. If the utility I’m recalling won’t do the trick, then there’s probably some code to write and compile, shouldn’t be too difficult.

What’s disappointing is that this is just display sleep, not system-wide sleep. In other words, everything is running except for the displays, so the logging function should still be working (and it is: there are things that are recorded in the log, such as the maintenance tasks scheduled for 3 am).

I might explore these additional logging options you mention. Thanks for the tip.

This is a backwards way of suggesting that you might want to simplify as much as possible. I _would_ try working with only one display, for example.

I’m to the point of speculating that Apple has simply not every tried to figure out the contradiction of a Unix system that can be put to sleep. I imagine they have no easy way of accurately determining how users actually use their systems in this respect, so they gave up gave up trying to guess, and decided to “see what happens”.

And with absolutely no evidence whatsoever –just a gut feeling– I’m going to guess there are some Unix processes that were writen with the assumption that they would never be interrupted. That’s true — they aren’t ever interrupted on a “traditional” Unix box. If we put our systems to sleep or wake up at some very narrowly defined places — Kablooie. Now, we just have to find those places.

I can also try to work with only one display for a while. That would be more frustrating (I usually need all the space), but I might try it if all else fails.

I agree with what you say about Unix and sleep issues. It’s probably quite true. But display sleep shouldn’t have any impact, since Unix is also designed to work just fine on a “headless” machine. The status of the display itself should be indifferent. Maybe Apple just hasn’t figured out how to treat display sleep features separately from system-wide sleep issues.

Display sleep _should_ not be an issue, as you say. But there’s that Mac functionality grafted on to Unix. Hmmmm, maybe we’re seeing two different issues: One is specific to display sleep and causes freezes when a transition occurs at a vulnerable time. Another is specific to system sleep.

Are you using dial-up or cable/DSL? I’ve got Entourage checking about 5 accounts via DSL, and I have noted some strangeness around sleeping and waking up in the midst of external ISP accesses. No crashes, but some rather alarming hesitations. I would imagine it would be worse with dial-up.

Pierre, my suggestion, while being a pain, would be to close Mail whenever you leave the system. However, I don’t think that’s a real possibility. Do your logs give you any indication as to the last process that was running. Maybe something in there can give you a clue. Maybe you can find a pattern in what was supposed to be processed next.

Anyway, when it comes to display sleep, as far as I can tell, the freeze occurs at a time when there is no change in the sleep status. For example, based on the last e-mail messages I had in Mail when I restarted the machine, the last freeze (see this post) occurred after midnight last night, i.e. at least 2 full hours after I left the machine. The screen saver kicks in after 15 minutes and the display sleep after 20 minutes, so the displays were already fast asleep when this happened.

I am using dialup, but via an AirPort Base Station. I have noticed some flakiness in Mail in relation to the slow modem connection, but no freeze.

Freezes: Just for fun, disable your screen saver and display sleep entirely. If one of your monitors is a CRT, power it when you’re done each evening. See what happens.

But that’s probably not going to make any difference.

Here’s a thought. When you’re done each evening, open a command window and type

% top -l 0 > reallybiglogfile

which might save something useful about what was going on near the end. You can simply delete the log file if nothing happened.

Second thought: this probably won’t work. You need a variant that executes often –say, once a second– and closes the log file after each output. This will look something like this as a script:

…
rm biglogfile
Begin
sleep 1
top -l 1 >> biglogfile
Repeat
…

but the exact form of the endless loop will depend on the shell you are using. I could look into this a bit more — I’m not an instant shell script writer.

What you want to do is have top tell you the “most” active process within 1 second of your actual crash. This method ought to catch the culprit, gui or daemon.

Before you do that –I think someone suggested this already– try quitting mail.app before you quit for the evening and see if this fixes the issue. That would be my best guess at this point.

FYI, I am using the default shell in Panther, i.e. bash.

the following is a copy/paste from a bug i submitted to apple:

03-May-2004 04:20 PM Nate Friedman:
Summary:
Locking the screen and locking the keychain may cause the computer to get stuck when waking from sleep.

Steps to reproduce the problem:
1. Set the keychain to lock when you put the computer to sleep
2. Set the computer to lock the screen when you put the computer to sleep
3. Run a program that accesses the keychain on a timer (I use Microsoft Entourage X)
4. Put the computer to sleep
5. Wake the computer to sleep. Do not unlock the screen
6. Wait until you know that your timer has gone off to try to access the keychain
7. Put the computer back to sleep
8. Wake the computer

Alternate Step 5-6 (unverified)
5. Wake the computer from sleep
5a. Unlock the screen
6. Wait until the timer goes off to try to access the keychain. Do not unlock the keychain

Expected Result:
I would expect it to prompt me for a password to unlock the screen

Actual Result:
The computer wakes to a black screen with a busy curser. The unlock prompt never appears.

My guess as to what is going on is that the security server has blocked waiting for me to unlock the keychain. Unfortunately, i can’t unlock it because i can’t get to the password prompt!

This issues seems to happen to me around once every 2 weeks, and aside for not being sure if i’ve unlocked the screen, this is the only situation that forces me to 3 finger solute my machine. (i’m sure i could SSH in, to cleanly reboot it, but that’s even more work)

Workarounds:
Don’t try to access the keychain on a schedule near the time you plan to wake the computer.
or
Don’t set the keychain to lock on sleep.
or
Don’t set the computer to lock the screen on sleep.

Isolation:
This happens both on my old PowerBook G4 and new iBook G3

Note:
I’ve been having the problem since 10.3.0, but i didn’t know about the bug reporter until now.

Disabled the screen saver. Fine for now.

Mare6