Subject: Re: Volunteer(s) needed to test mvme147 lpt driver
To: None <thorpej@nas.nasa.gov>
From: Steve Woodford <steve@mctavish.demon.co.uk>
List: port-mvme68k
Date: 07/15/1996 21:29:41
Jason Thorpe wrote:
> Anyhow, I do have some stylistic/architectural suggestions...
>
> In lptreg.h, don't declare a global "LPT" pointer to the register union.
> In the places where's it used, you have the softc available. Probably
> better to just carry the pointer to the registers in the softc and
> reference them that way. No functional difference, just slightly better
> style. It's a nit :-)
Point taken. I rattled this driver up just this morning, so there's
bound to be some rough edges!
> Also, every device with a devsw entry should get a "needs-flag". In the
> case of the lpt driver, you really need to use NLPT in the cdevsw entry.
> "needs-flag" just generates a "1" or "0" based on whether or not the
> driver is included in the kernel. The macros that expand to the switch
> table rely on this number being accurate. With the conf.c below, a
> kernel without the lpt driver will fail to link because the "1" forces
> the macros to generate references to the non-existent driver entry points.
Yup. I noted this myself when I tried to recompile a kernel without lpt.
I emailed chuck to include the needs-flag change if he adds lpt to the
main tree.
> Anyhow, these are just trivial little nits... I think it's totally cool
> that you did this :-)
<grin> Amazing what you can do if you take a day off work sick :)
Cheers, Steve