0.5.5 release planning (Was Re: [PATCH] fix spec if rebuild from distributed package)

[Jon Nettleton]

On Tue, 2005-11-01 at 17:03 +0000, Richard Hughes wrote:
> On Mon, 2005-10-31 at 12:44 -0500, Jon Nettleton wrote:
> > On Mon, 2005-10-31 at 12:31 -0500, David Zeuthen wrote:
> > > On Mon, 2005-10-31 at 17:05 +0000, Richard Hughes wrote:
> > > > David, When's 0.5.5 going to be forked? There's lots of goodness in
> > > > CVS not in 0.5.4
> > > 
> > > I don't see any problems with doing a 0.5.5 release soon. Maybe just
> > > wait for Jon Nettlon's bugs to be worked out though? It's also mostly
> > > about what you and Danny think since I'm not too involved these days.
> > > 
> > >     David
> > > 
> > > 
> > > _______________________________________________
> > > hal mailing list
> > > hal at lists.freedesktop.org
> > > http://lists.freedesktop.org/mailman/listinfo/hal
> > 
> > For a 0.5.5 release I am happy with the existing fixes we have worked
> > on.  I think we are 95% to getting removable media working consistently,
> > and reliably on the desktop ( at least from a Hal perspective ).
> > 
> > Has anyone had a chance to look at the last patch I submitted as a
> > re-work for the race condition on /proc/mounts?  I am curious if you
> > think that it is just over-complicating the process.
> 
> It does seem very complicated.. in little words (for simple me), what
> happens when the race happens, and is this a kernel bug?

I believe that in the earlier thread when the first patch was proposed
by Danny Kukawka, there was some discussion as to whether this was a
kernel bug or not.  In that thread Kay Sievers seemed to be of the
opinion this was the proper way to work with the kernel, and it's
possible event timing idiosyncrasies.

> 
> Does the mounting work now?

With Danny's patch the mounting did work much better, almost perfect.  I
created my revision for the following reasons.

1)  Under high latency environments this gives a bit more stability by
retrying more and giving up to .6 seconds for /proc/mounts to be updated
2)  It stays in the function for shortest possible time (relatively) for
both mount and unmount events. Previously unmount had to always go
through the delay loop.
3)  By doing obscene things on the desktop, ie. mounting and unmounting
a cdrom while unpugging my 6 port card reader with multiple cards in it
and plugging and unplugging multiple usb drives I could get the race
condition to happen.  I know this is an absolutely ridiculous situation,
but I would like the removable media process in Hal to be as reliable as
possible.

> 
> As for 0.5.5, my acpi, apm and pmu updates are always trickling in, so
> whenever is convenient for you David, is good for me.
> 
> We haven't had a release in ages, and I'm fed up of telling the ubuntu
> and rawhide people "bug fixed in CVS"... :-)
> 
> Richard.
> 

I am not in love with this exact patch by any means, it was more of a
design proposal.  I am completely open on any changes suggested, or
ditching it all together.

Thanks for your time.

Jon