[Usability] Re: Bluescreens

[michael chang <thenewme91 gmail com>]

On 9/22/05, usability-request gnome org <usability-request gnome org> wrote:

> One of my HDDs died some days ago, in the middle of a GNOME session.
>
> This, of course, provoked some error in the kernel SCSI driver (or
> something), and the computer froze. I could change TTY and tried to
> log in at the prompt, but the keyboard didn't work.
>
> I just think it'd be nice if GNOME fired up a "bluescreen" saying

IIRC, KDE fires up a error box when a program dies weirdly, but... (see below).

> something along the lines of
> "Something serious just happened. Please call someone who knows what
> they're doing (presumably your vendor or your grandson). Restarting
> the computer might help, but don't count on it.

Actually, if you have a decent journaling file system, in many cases
where the machine freezes, restarting the computer fixes it.  And, if
you compiled SysRq hacks into the kernel (usually no), you can push
Alt-SysRq-B to reboot the machine, I think.  [Yes, it's a soft reboot
but not a proper shutdown, and it saves ~3 seconds you use waiting
holding the power button waiting for the power-off trigger to go off.]

> Warning: Technical mumbo-jumbo follows. You might want to make a note
> of the most common words in the output below (if a word is repeated
> many times) and relay them to the person you're about to call.
>
> <kernel output>"

AFAIK, this is not technically possible, if the kernel panics, then
there is no way to send output to the X screen, or at all -- simply
because X processing stops, and all processing stops.  You can't even
capture Kernel Panics' outputs.  And, in theory, you shouldn't be able
to.  Too dangerous.

For example, What happens when the bug is in the display driver?  E.g.
I start up my display driver, triggers kernel bug, gets captured, so
GNOME tries to open the display to show the error, triggers bug again,
and then the computer is trapped in an infinite loop until it runs out
of memory.

> I just think it's better than freezing up with no indication of what's

Linux will also run on slower machines and/or slower hard drives, when
the CPU is waiting for a huge write, or for the CPU, it may appear to
freeze - we don't want to accidentally think an app is frozen when
it's just the CPU is a bit slow, and the user doesn't mind waiting. 
XP made this mistake with it's apps.  If, in any case, a Application
is frozen, I should always restart my computer (although if the rest
of the system responds, I do this nicely) -- no exceptions.  Simple as
that.

Also, X may "freeze" and the console could be fully operable.  That's
happened to me a lot.

I don't know about other people, though.

> wrong - or that there is something wrong at all.

When an app dies, sure.  Otherwise... well, then you have to go into
writing an error handler for the Linux Kernel.  This isn't something
that GNOME is capable of doing, AFAIK.

And IMO, the BSOD was the worse trait of Windows 9x.  Everyone I know
who is a consumer thought that it should not have even existed.

And personally, you simply shouldn't be triggering kernel bugs in the
first place.  If a SCSI drive fails, it shouldn't prevent you from
restarting the rest of the system, just so long as you don't attempt
to mount it; and the rest of it should be done by a professional doing
drive recovery.  You shouldn't even be using drives that are likely to
fail in the first place (although this doesn't apply if your hard
drive failed "spontaneously".)

--
~Mike
 - Just my two cents
 - No man is an island, and no man is unable.