[[ David - CC'd to you in-case there's anything that may be useful
   in here to put on the www pages somewhere :-) ]]

Richard Michael wrote:

> I noticed the conversation between Simon and Dan regarding
> update to the sys/dev/tc/asc_ioasic.c file - specifically
> updating from rev. 1.15 to 1.15.2.1 and after looking on 
> the ftp site, I am somewhat confused about the whole matter.
> 
> I just built a new kernel from:
> /pub/NetBSD/NetBSD-1.4.1/source/sets/syssrc.tgz
> 
> but upon inspection I see it contains the old asc .c file.

Correct - this file contains the source code at the time of the 1.4.1
release.  It will never change.

> Since,
> /pub/NetBSD/NetBSD-current/tar_files/src/sys.tar.gz (25/12 @ 13:45)
> and
> /pub/NetBSD/arch/pmax/snapshot/1.4.2_ALPHA/source/sets/src.tgz
> 						(27/12 @ 12:00)
> 
> are both development (the latter is a 1.4.2 ALPHA, what about
> NetBSD-current?), which should I use for a bleeding edge kernel?
> (Not necessarily a stable one.)

This crops up a lot :-(

I _thought_ we had a description on a web page somewhere but I can't
place it right now.  In a nutshell, 1.4<letter> (eg, 1.4A, 1.4P) is
the "bleeding edge" and this is called "NetBSD-current".  1.4.<number>
is "stable" with well proven patches and non-significant changes, but
should functionally be little different to 1.4.  Upgrading a 1.4 machine
to 1.4.1 or NetBSD-release shouldn't cause any user-visible changes.
1.4.2_ALPHA should me more stable (and more secure because it includes
the latest BIND code) than both 1.4 and 1.4.1.

So:

	1.4   -->   1.4.1
	1.4A        1.4.2
	1.4B
	 ..
	1.4P

Developement (the "bleeding edge") continues down the first column,
while bugs and security fixes are added to the second column.
Eventually the first column will turn into 1.5 or 2.0 (whichever version
number we choose to move to next time there's a major release).

So the "bleeding edge" is pub/NetBSD/NetBSD-current and the "stable" is
in pub/NetBSD/NetBSD-release.


One more thing - I believe Dan was working with 1.4.1, so his
NetBSD-release version of sys/dev/tc/asc_ioasic.c would need to be at
revision 1.15.2.1 to include the DMA fix.  If he were using NetBSD
current, he would need to be using rev 1.18 or later.  From this, we can
see that 1.4.1 is based on rev 1.15 of the file, while the "bleeding
edge" is up to rev 1.18.  The differences between 1.17 and 1.18 (the DMA
fix) were added to rev 1.15 creating 1.15.2.1.  Hmm, does this paragraph
help?

I really hope the above is as clear as mud!

> Also, there's this file:
> /pub/NetBSD/NetBSD-release/tar_files/src/sys.tar.gz
> 
> Since this a daily snapshot of the release branch, should
> I assume that it is of the 1.4.1 release?  Why is there a
> 1.4.2 changelog in the directory if 1.4.2 is only alpha?

This is the "1.4 branch" or "stable" to use another term.  The 1.4.2
changelog is a list of what has changed since 1.4.1.  When 1.4.2 is
released, we'll then maintain a 1.4.3 changelog which will then list
everything that has changed since the 1.4.2 release.

> I see that there are no patches in:
> /pub/NetBSD/arch/pmax/patches/

> To what source tree would these patches be relevant?

This directory contains random patches.  Hopefully a README file would
describe what they are applicable to.

> What source set should I use for the latest stable kernel?
> I would guess the NetBSD-release source, since I notice it
> contains the asc_ioasic.c 1.15.2.1.

Yup!

> --
> Another issue:
> For the build of the 1.4.1 kernel that I mentioned above,
> I commented out this line in my CONFIG file:
> 
> #pseudo-device   fb               1      # up to 3 framebuffers
> 
> since I don't have a framebuffer, nor will I need one.  The build
> failed errors in the link stage.  I don't have them copied out,
> but would be happy to reproduce them.  My question is just why
> it failed.  Is there something I mis-cfg'd (does something else also
> need to be commented out) ?  Once I uncommented the line, kernel
> built fine.

I'd guess you also need to comment out the following lines to (from the
-current GENERIC config file):

	mfb*    at      tc? slot ? offset ?     # PMAG-A MX Monochrome Framebuffer
	cfb*    at      tc? slot ? offset ?     # PMAG-B CX Color Framebuffer
	sfb*    at      tc? slot ? offset ?     # PMAGB-B HX Smart Framebuffer
	px0     at      tc? slot ? offset ?     # PMAG-C PX family
	xcfb*   at      tc? slot ? offset ?     # PMAG-DV Color Framebuffer at MAXINE
	pm*     at      ibus0 addr ?            # 3100 onboard fb

	pseudo-device   rasterconsole    1      # NB: raster console requires "fb"

All of these relate to framebuffer type things.  If you still have
problems after this, please include your error output and that should
help narrow things down.

Simon.