Re: oops pauser. / boot_delayer

[Josef Sipek <jsipek@fsl.cs.sunysb.edu>]

On Mon, Jan 09, 2006 at 12:08:27AM +0100, Pavel Machek wrote:
> On Ne 08-01-06 14:30:00, Josef Sipek wrote:
> > On Sun, Jan 08, 2006 at 02:21:32PM +0100, Pavel Machek wrote:
> > > On Pá 06-01-06 00:36:09, Dave Jones wrote:
> > > > So disable CONFIG_CRYPTO_TEST. There's no reason to test this stuff every boot.
> > > 
> > > Maybe even with CRYPTO_TEST enabled we could only report _failures_?
> > 
> > Why? As far as I know, it is intended for developers as a regression test. I say
> > if you don't like the output, make the thing a module or don't compile it at all.
> 
> I don't like the output, but if it only reported failures, I could
> leave it running and potentially catch some strange failures.

I agree that it is useful to know about strange failures, however I still maintain
that _if_ the module is intended as a regression test for developers, than the
excessive (?) output is fair. I think that the most logical course of action is to
have a verbosity module paramter which defaults to displaying errors only, but it still
allows developers to get all the information they need.

> Is reporting successes actually useful?

Then I propose: :)


diff -r b4fca0ece97f kernel/sys.c
--- a/kernel/sys.c	Sat Oct 22 19:24:10 2005 +0300
+++ b/kernel/sys.c	Sun Jan  8 18:26:49 2006 -0500
@@ -436,7 +436,6 @@
 void kernel_halt(void)
 {
 	kernel_halt_prepare();
-	printk(KERN_EMERG "System halted.\n");
 	machine_halt();
 }
 EXPORT_SYMBOL_GPL(kernel_halt);

Jeff.