Re: CONFIG_PCI_BIOS in 2.1.132

Michael Elizabeth Chastain (mec@shout.net)
Tue, 29 Dec 1998 18:08:58 -0600


Hi Riley,

> Attached is the .config file produced by "make menuconfig" in
> 2.1.132ac13 - you will note that the output produced does indeed show
> comments for the "not set" options...
>
> # CONFIG_PCI_GOBIOS is not set
> # CONFIG_PCI_GODIRECT is not set
> CONFIG_PCI_GOANY=y

Oh, those are not comments! They are semantically meaningful lines
that happen to start with '#'. All the flavors of configuration
translate those on input to "CONFIG_PCI_GOBIOS=n', et cetera.

By the way, the reason these are comments is because the Makefiles
read .config, and these *are* comments to the Makefiles. And the
Makefiles do care about the difference between something with a null
value and something with a value of "=n".

I didn't design this configuration system. I hate this configuration
system. I am not proud to be the maintainer of Menuconfig, even though
I asked for the job. The very existence of xconfig causes me pain.

But the part of the job I hate the most is when everyone else in the
world with a text editor writes untested crap into Config.in files and
then says "oh just make all flavors of Configuration smarter please."
I CAN'T KEEP UP WITH THIS. And it chews up time that I want to use to
rewrite this whole compost heap of code the right way.

> That "unset" line was inserted by Linus, not me - and, as per the
> missive he posted to you in the linux-kernel list, it IS necessary, so
> it's "make xconfig" that's in the wrong here...

Actually, I don't agree with that. But I respect Linus's judgement
even when I think he's wrong.

So now I am hip deep in kconfig.tk trying to figure out how it manages
its symbols. And then I can't simply edit code, I have to change tkgen.c
and tkparse.c to recognize the 'unset' statement and *emit* the right
thing in kconfig.tk.

I am probably going to suggest the following patch instead:

+ else
+ define_bool CONFIG_PCI_BIOS n
+ ...
+ else
+ define_bool CONFIG_PCI_DIRECT n

mec> We do need help text, but I don't think the help text needs to talk
mec> about the internal options.

rhw> Again, that was Linus' suggestion...I tend to bow down to experts like
rhw> him...

Well, I won't argue with that one.

Michael Elizabeth Chastain
<mailto:mec@shout.net>
"love without fear"

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/