Re: CVS commit: [itohy-usb1] src/sys/dev/usb

[Marc Balmer <marc%msys.ch@localhost>]

* M. Warner Losh wrote:
> In message: <19578.1182847028%dtsp.co.nz@localhost>
>             david.sainty%dtsp.co.nz@localhost writes:
> : "M. Warner Losh" writes:
> : 
> : > In message: <10105.1182771473%dtsp.co.nz@localhost>
> : >             david.sainty%dtsp.co.nz@localhost writes:
> : > : Brilliant!  I have been working on this, and throwing up my hands in
> : > : horror over the slapstick OpenBSD changes!  Not to mention the nasty
> : > : state of the device definitions added for this driver in the OpenBSD
> : > : usbdevs (dups, definitions all over the place).
> : > 
> : > The BSDs should just share one usbdevs.  Having multiple ones is just
> : > silly.
> : 
> : But it should be the NetBSD one, I'm pretty sure the FreeBSD one had
> : copies of the same mistakes when I looked :)
> 
> It shouldn't be the NetBSD because it is missing a number of entries
> that are in the FreeBSD one.  Also, its entries are purposely
> truncated: bug or feature is debatable.
> 
> It shouldn't be the FreeBSD one because it is missing a number of
> entries in the OpenBSD one.
> 
> It shouldn't be the OpenBSD one because it is missing a number of
> entries in the NetBSD one.
> 
> Etc.
> 
> I've also fixed the FreeBSD one to fix the mistakes (like the NetBSD
> one gets fixed from time to time).  I've fixed this mistakes in the
> NetBSD one as well.  Ditto to the OpenBSD one.  I've been doing this
> for years.
> 
> At this point, the one I have in FreeBSD is close to being the best
> superset to them all.  The biggest difference is that NetBSD's uses
> shorter names, sometimes arbitrarily so, for many of the company and
> device names.  To effectively share, we need to figure out how to
> generate the shorter names from the longer names.  NetBSD did this for
> kernel space reasons since it compiles this file into the kernel.
> FreeBSD doesn't, by default.  Ideally, if one group wants to subset
> the file for inclusion into the tree, then that subsetting should be
> done mechanically from the common file.
> 
> I have uncommitted patches in my tree that parse this file and report
> via usbdevs any matches that are found in this file in a verbose and
> useful way.
> 
> The bottom line is that there needs to be a *MERGED* version, and that
> *MERGED* version will likely have code impacts in *ALL* the BSDs.
> I'll do the leg-work in FreeBSD and NetBSD, but someone needs to
> integrate into OpenBSD.  I'll also let others do the NetBSD work too,
> I'm not proud.

Just a short sign of live...  I am discussing this on the OpenBSD side,
but one person involved has it's exams right now and is a bit short on
time.

> 
> Warner