[9fans] Booting on slow CPUs

[9fans] Booting on slow CPUs

[Lucio De Re]

I seem to have isolated the problem to a race (almost literally)
condition between init0() terminating and whatever may be happening
behind the scenes that seems to beat init0() and cause it to stall.

If I empty init0()'s inner core (the bit controlled by !waserror()),
the booting procedure procceds, however crippled.  Various
strategically placed "print" calls give different degrees of progress.

If anybody can save me much further searching to find out which
operation is trashing init0(), I would appreciate the help greatly.

A question I have to ask, is whether poperror() within a section
controlled by !waserror() isn't in the wrong place.  But I don't for
one moment think this has any bearing on my quest.

++L

Re: [9fans] Booting on slow CPUs

[Lucio De Re]

On Sat, Mar 16, 2002 at 01:55:12PM -0500, Russ Cox wrote:
> 
> Sorry, I thought you'd looked at waserror() before
> and had just missed this one case.  It's not a normal
> function.  It's used to implement exceptions via
> a setjmp/longjmp-equivalent and a stack of jmpbufs.  
> 
It's a reasonable assumption, I have to confess.  It's a difficult bad
habit to break, this shooting from the hip.  I forget that in this
forum I'm in the lower echelons, unlike my daily surroundings :-)

> See http://www.vitanuova.com/inferno/man/10/error.html.
> 
Thank you, I hope that will help.

++L

Re: [9fans] Booting on slow CPUs

[Russ Cox]

Sorry, I thought you'd looked at waserror() before
and had just missed this one case.  It's not a normal
function.  It's used to implement exceptions via
a setjmp/longjmp-equivalent and a stack of jmpbufs.  

See http://www.vitanuova.com/inferno/man/10/error.html.

Russ

Re: [9fans] Booting on slow CPUs

[Lucio De Re]

On Fri, Mar 15, 2002 at 03:07:49PM -0500, Russ Cox wrote:
> 
> You need to poperror() on the path where there's no error, thus:
> 
> 	if(waserror()){
> 		/* no poperror on this path ... */
> 		return nil;
> 	}
> 	poperror();
> 
I guess I ought to look at waserror(), but surely the above is totally
counter-intuitive?  Maybe a poor selection of function names?

> but
> 
> 	if(!waserror()){
> 		/* do some stuff... */
> 		poperror();
> 	}
> 	/* more stuff */
> 
> The second is idiomatic for ``do some stuff, stopping early
> if you encounter an error; either way, then do more stuff.''
> 
Sorry, I'm being obtuse here, but perhaps you want to rephrase your
description: would

"if no error, do some stuff; irrespective, thereafter do more stuff"

be more accurate (if slightly mangled English)?

++L

Re: [9fans] Booting on slow CPUs

[Russ Cox]

You need to poperror() on the path where there's no error, thus:

	if(waserror()){
		/* no poperror on this path ... */
		return nil;
	}
	poperror();

but

	if(!waserror()){
		/* do some stuff... */
		poperror();
	}
	/* more stuff */

The second is idiomatic for ``do some stuff, stopping early
if you encounter an error; either way, then do more stuff.''

Russ