[Yaird-devel] Bug#345067: ide-geenric inclusion even if it doesn't exist.

[Sven Luther]

On Sun, Jan 29, 2006 at 01:09:13PM +0000, Stephen Gran wrote:
> This one time, at band camp, Sven Luther said:
> > Hi Erik, ...
> > 
> > I hope you have been fine, passed well into the  new year and in
> > general had a good start of the year.
> > 
> > I am writing this to ask you to have a look at the patch in :
> > 
> >   http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=345067
> 
> Well, as you pointed out, this patch seems like a bad hack.  You're just
> special casing one known broken architecture, rather than actually

Well, it has been broken since over a month now, and means the kernel is
uninstallable with yaird, so i believe there is some kind of urgency involved
here, so why not please appply this patch as a workaround ASAP, until a proper
solution can be found ? 

> addressing the real problem.  The real problem is that yaird fails when
> it tries to modprobe a module that doesn't exist.

I thought so at first, but after more thinking about the yaird philosophy, i
don't think this is a bug but a feature. If the module doesn't exist, then it
should well be that the system will fail to boot because of it.

This case is different, because not only the module does not exist, but it is
also not really needed, and my patch is only to not mark it as needed like it
is now, because i, wearing my powerpc debian kernel maintainer hat, know that
it is not needed on powerpc.

Actually, i believe it is not needed on i386 too, and that this stuff done
here is only a workaround for a real solution involving fixing the
via/piix/whatever modules.

So, this is a fix t oa hacky workaround already implemented in the first
place.

> If it is reasonable to move on when a module that yaird thinks should be
> loaded does not exist, then it seems to me that the better way to fix
> this would be in Modprobe.pm and Base.pm.  First have modprobe.pm run
> modprobe -l $module, and if !$rc, return something special (0? -1?,
> "no-module"? whatever - just so it's not the same return code that
> something else uses).  Then notice that case in the exec bit in Base.pm,
> and don't raise a fatal error.

Maybe a --ignore-missing-modules option should do this.

> If that assumption is unreasonable (I honestly haen't looked at yaird
> enough to know - I can imagine arguments both ways), then the fix would
> be adding ide-generic to the module blacklist on the architectures where
> it causes a problem.

Huh ? i am not familiar enough with yaird to know about this blacklist, but i
believe that having Hardware.pm force-include it may override any kind of
blacklist, and should be done with more discrimination in the first place.

> It also seems to me that the best way to fix the via82cxxx module
> needing ide-generic but not depending on it would also be in
> Modprobe.pm.  There, you can just add 'insmod ide-generic' to @lines,
> and it also seems the right place to test for it.

Possible, but this is the hack which already went hin, so what can i say.

> But again, I haven't looked at yaird enough to know if I'm missing
> something.

Thanks for your feedback.

/me just hopes we don't need a multi-month flamewar about this before thigns
get fixed :/

Friendly,

Sven Luther