hangup on bootscripterror in shutdown

[sebastianmeisel@…]

When for some reason I get an error message from a bootscriptduring shutdown my computer hangsup totally. I follow the instruction: Press any key to progress...but there is no reaction and I have to turn the computer off.I'm running LFS on a IntelPentium III mobile Notebook.All programm's mentioned in the book are compiled without optimation flags.

[gerard@…]

I've seen the same problem here. It doesn't happen every time (i've rebooted a whole bunch of times to try and reproduce it, but I couldn't find a pattern in it's behaviour). When it does and I hit enter to continue, I get ^{M on the screen and it won't continue. During my last 'frenzy' I started hitting all kinds of key combinations and all of the sudden it did continue. Unfortunately I don't remember what keystroke I hit to get it to continue (that would give a clue what's going on). We'll keep an eye on it but it's a tough one to track down when there is no easy way to reproduce it.}

Kernel glitch perhaps?

[gimli@…]

^Q?

[gerard@…]

No if the terminal would be paused you wouldn't be able to type anything. When it happens (hasn't happened for a few days now) I can type all characters except the enter key is displayed at ^{M instead of literally} M execution.

[gerard@…]

Going to be tough to find out, so let's not spend too much time on it. We'll keep it open until we foudn the problem. Even though it's marked P1 for LFS-4.0, if it can't be figured out, P5 it later on.

[timothy@…]

I experienced this the other day. IMO, the 'read' line should be replaced with 'sleep 10' or something, and just log the error someplace. Having to hit enter from the machine feels too much like windoze.

[timothy@…]

Again, I recommend that the "read" line in functions be replaced with 'sleep 10' or something similar. When I ran into this problem, I was far from home, so I had to call my parents to hit the enter key :) IMO, that is totally unacceptable for a server.

I think the error should be logged instead.

[gerard@…]

True not great for a server. Then just make changes when you install a server. The idea is that you may not want to boot a system when some bootscript(s) fail. This if for the user to decide, I don't think it's appropriate as a default behaviour.

[bryan@…]

Old bug I know, but I just saw it mentioned on lfs-dev and figured I'd take a look.

Gerard, you hit Ctrl-J to get "read" to finish, I think.

What your terminal shows as ^{M is ASCII code 13 (and M is the 13th letter, which is why that matches up that way), which is carriage return. "read" is looking for newline, though, which is ASCII 10 (and J is the 10th letter, so you can input an ASCII 10 by hitting Ctrl-J).}

Plus, I've run into this once too, and Ctrl-J bypassed the prompt. ;-)

(Perhaps it would make sense to change the bootscript prompt "press enter to continue" into "press enter to continue, or Ctrl-J if that doesn't work"? It won't fix the problem, but at least it'll give a clue what to do...)

I'll also check in a minute whether it happens if I create a dummy shutdown script that just does an "exit 1" (to test whether it happens every time a shutdown script "fails"), but I'll have to shut down to do it. Hang on.

[bryan@…]

Never mind, I can't seem to reproduce it reliably either.

Across about 10 reboots, it happened once, and only then after using the system for a while (couple of VCs had been in use, and a couple of instances of vim 6.1 had been open too, editing some of the bootscripts). Might possibly be something related to vim or VCs or the escape sequences used to change the color, but I don't really know for sure.

I could never get it to happen on LFS-CVS-20030515 (with lfs-bootscripts-1.11), FWIW; the one time it happened was on my "LFS 3.3+updated packages" partition. But I also had a bunch of reboots (with the dummy "exit 1" script in place) work OK on that partition.

So much for trying to come up with a cause... but at least there's a workaround if it does happen.