Re: [MiNT] SIGPWR

[Guido Flohr <gufl0000@stud.uni-sb.de>]

Hi Thomas,

On Tue, Oct 19, 1999 at 03:50:29PM +0200, Thomas Binder wrote:
> Hi!
> 
> As promised, I looked at the problem Guido had adding a new signal to
> signals.h of the MiNT source. The reason the resulting kernel crashed
> was a missing dependency in the Makefile: genmagic must be recompiled
> when signals.h has changed, as changing NSIG in there results in
> different offsets in the PROC structure, and magic.i has to have the
> correct ones.

Praise and glory!  Finally ...

I already thought I was nuts.  The first time I did the change the kernel
worked just fine, I was quite sure.  But I must have changed signals.h
before I compiled genmagic, that explains it.

> 
> Thus, to add SIGPWR as signal no. 31, simply add it to the list of
> #defines in signals.h, change NSIG to 32 in the same file, and add
> include/signals.h to the list of dependencies for genmagic.ttp in the
> Makefile.

To avoid similar problems in the future I will automatically generate the
dependencies for genmagic (don't worry, you can still add more
dependencies for the target in the actual Makefile).

> 
> Note that there's no way to further increase the number of available
> signals using this method, as otherwise the number of signals would be
> greater than 32, forcing a change in the signal masks, which are longs
> and can't therefore handle more than 32 signals (1 bit per signal).

That's clear.

I conclude: Nobody really complained about SIGPWR in general, correct?
It is ok to add it as #31 and we recess the SIGSYS discussion sine die
(resp. to the day we miss the next signal).  Also correct?

Thank you, Thomas! That was quick.

Ciao

Guido
-- 
http://stud.uni-sb.de/~gufl0000/
mailto:gufl0000@stud.uni-sb.de