Well, time for my weekly bug-report again. Nothing too bad this time.

1) linux/kernel/signal.c kills a process with the wrong error code when
it gets a signal that the process doesn't handle.  The easiest way to
test this is by sending a SIGQUIT to some running program (with ^|) -
bash will report that the program died due to "Illegal Instruction"

This bug already got corrected in some version, but it wiggled its way
back :-(.  The line is something like this (in do_signal()):

	if (!sa_handler) {
		if (nr == SIGCHLD)
			return;
		else
			do_exit(1<<(nr-1));
	}

when it should be

	if (!sa_handler) {
		if ...
			...
		else
			do_exit(nr);
	}

2) As has been shown, you can get spurious "out of memory" errors if the
passwd file has the incorrect format.  This wasn't a error in the
kernel, but in the library - I'll correct it by using the GNU lib etc. 
I'll also make full library source available (not before sunday: I've
got 1e3 things to do).  In the meanwhile, this bug isn't easy to
trigger, and not fatal even then, so... 

		Linus

PS.  I'm interested in hearing about everybody that has problems with
either the harddisk or the serial line.  There was a report that serial
lines sent out 2 characters at a time - does anybody else see this?