config language shortcomings in 2.4

config language shortcomings in 2.4

[Joshua Kwan]

[-- Attachment #1: Type: text/plain, Size: 1539 bytes --]

Hello,

Recently the Debian kernel team has been discussing 2.4's configuration
system. We have a patched tg3.c which removes the firmware and adds a
dependency on CONFIG_FW_LOADER for cases where operation is impossible
without the firmware.

Initially, we changed this:

dep_tristate 'Broadcom Tigon3 support' CONFIG_TIGON3 $CONFIG_PCI

to this:

dep_tristate 'Broadcom Tigon3 support' CONFIG_TIGON3 $CONFIG_PCI $CONFIG_FW_LOADER

But this would still be selectable if CONFIG_FW_LOADER were not set at
all. Selecting it would not enable CONFIG_FW_LOADER either.

So, we tried this:

if [ "$CONFIG_FW_LOADER" != "n" -a "$CONFIG_EXPERIMENTAL" = "y" ]; then
   dep_tristate 'Broadcom Tigon3 support' CONFIG_TIGON3 $CONFIG_PCI $CONFIG_FW_LOADER
fi

and this STILL doesn't seem to work in the case that CONFIG_FW_LOADER
isn't set but CONFIG_EXPERIMENTAL is y.

Eventually we continued droning through the corner cases until reaching

if [ "$CONFIG_EXPERIMENTAL" = "y" -a \
     "$CONFIG_HOTPLUG" = "y" -a \
     "$CONFIG_FW_LOADER" = "y" -o "$CONFIG_FW_LOADER" = "m" -a \
     "$CONFIG_CRC32" = "y" -o "$CONFIG_CRC32" = "m" ]; then
       dep_tristate 'Broadcom Tigon3 support' CONFIG_TIGON3 $CONFIG_PCI $CONFIG_FW_LOADER $CONFIG_CRC32
fi

which finally has the desired effect.

But is there really no way to say

config TIGON3
       select FW_LOADER

in 2.4?

CONFIG_PCMCIA_ATMEL also suffers from this problem.

Any ideas? Is it just that the old Configure scripts suck?

-- 
Joshua Kwan

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 881 bytes --]

Re: config language shortcomings in 2.4

[Willy Tarreau]

Hi,

On Thu, Aug 19, 2004 at 12:12:29AM -0700, Joshua Kwan wrote:
 
> Eventually we continued droning through the corner cases until reaching
> 
> if [ "$CONFIG_EXPERIMENTAL" = "y" -a \
>      "$CONFIG_HOTPLUG" = "y" -a \
>      "$CONFIG_FW_LOADER" = "y" -o "$CONFIG_FW_LOADER" = "m" -a \
>      "$CONFIG_CRC32" = "y" -o "$CONFIG_CRC32" = "m" ]; then
>        dep_tristate 'Broadcom Tigon3 support' CONFIG_TIGON3 $CONFIG_PCI $CONFIG_FW_LOADER $CONFIG_CRC32
> fi
> 
> which finally has the desired effect.

I'm surprized, because I think you have a precedence problem here. Your 'if'
condition will be true if either :
  CONFIG_CRC32 = m
or
  CONFIG_CRC32 = y and CONFIG_FW_LOADER = m
or
  CONFIG_FW_LOADER = y and CONFIG_HOTPLUG = y and CONFIG_EXPERIMENTAL = y

Anyway, I believe that you have no other choice due to the way dep_tristate
works. What would you expect it to do when it depends on 3 variables which
are respectively 'n', 'm' and 'y' ? Honnestly, without looking closer at its
implementation, I would not be able to give a valid response.

BTW, have you tried defining a temporary variable somewhere ? There are
portions of config where you see things such as :

if [ CONFIG_XX = "y" -o CONFIG_YY = "m" -a CONFIG_ZZ = "y" ]; then
   TEMP=y
fi
dep_tristate "cool feature" CONFIG_COOL $TEMP

Perhaps it could help you define complex combinations.

Regards,
Willy

Re: config language shortcomings in 2.4

[Joshua Kwan]

Willy Tarreau wrote:
> Anyway, I believe that you have no other choice due to the way dep_tristate
> works. What would you expect it to do when it depends on 3 variables which
> are respectively 'n', 'm' and 'y' ? Honnestly, without looking closer at its
> implementation, I would not be able to give a valid response.
> 
> BTW, have you tried defining a temporary variable somewhere ? There are
> portions of config where you see things such as :
> 
> if [ CONFIG_XX = "y" -o CONFIG_YY = "m" -a CONFIG_ZZ = "y" ]; then
>    TEMP=y
> fi
> dep_tristate "cool feature" CONFIG_COOL $TEMP
> 
> Perhaps it could help you define complex combinations.

Sorry for the belated response, but certainly it should not be necessary 
to do:

if [ "$CONFIG_FW_LOADER" = "m" -o "$CONFIG_FW_LOADER" = "y" ]; then
	HAVE_SOME_FW_LOADER=y
fi
if [ "$CONFIG_CRC32" = "m" -o "$CONFIG_CRC32" = "m" ]; then
	HAVE_SOME_CRC32=y
fi

if [ "$HAVE_SOME_CRC32" = "y" -a "$HAVE_SOME_FW_LOADER" = "y" ]; then
	dep_tristate 'Broadcom Tigon3 support' CONFIG_TIGON3 $CONFIG_PCI 
$HAVE_SOME_FW_LOADER $HAVE_SOME_CRC32
fi

just to enforce an explicit dependency on some form of CONFIG_CRC32 and 
CONFIG_FW_LOADER. This is necessary because "$CONFIG_CRC32" != "n" 
doesn't work the way you think it would.

Anyway, it's all very disgusting and I'm inclined to just ignore it and 
maybe some benevolent soul will one day port Kconfig back to 2.4.

-- 
Joshua Kwan

Re: config language shortcomings in 2.4

[Nigel Cunningham]

Hi.

On Tue, 2004-08-24 at 15:04, Joshua Kwan wrote:
> Sorry for the belated response, but certainly it should not be necessary 
> to do:
> 
> if [ "$CONFIG_FW_LOADER" = "m" -o "$CONFIG_FW_LOADER" = "y" ]; then
> 	HAVE_SOME_FW_LOADER=y
> fi
> if [ "$CONFIG_CRC32" = "m" -o "$CONFIG_CRC32" = "m" ]; then
> 	HAVE_SOME_CRC32=y
> fi
> 
> if [ "$HAVE_SOME_CRC32" = "y" -a "$HAVE_SOME_FW_LOADER" = "y" ]; then
> 	dep_tristate 'Broadcom Tigon3 support' CONFIG_TIGON3 $CONFIG_PCI 
> $HAVE_SOME_FW_LOADER $HAVE_SOME_CRC32
> fi

Just in case someone goes to copy and paste the above, there's a typo:
CONFIG_CRC32 = m or = m.

> Anyway, it's all very disgusting and I'm inclined to just ignore it and 
> maybe some benevolent soul will one day port Kconfig back to 2.4.

Hehe. I had to do something similar when I made suspend 2 able to be
compiled as modules. It could certainly be more elegant.

Nigel
-- 
Nigel Cunningham
Christian Reformed Church of Tuggeranong
PO Box 1004, Tuggeranong, ACT 2901

Many today claim to be tolerant. But true tolerance can cope with others
being intolerant.

Re: config language shortcomings in 2.4

[Marcelo Tosatti]

> Anyway, it's all very disgusting and I'm inclined to just ignore it and 
> maybe some benevolent soul will one day port Kconfig back to 2.4.

Not worth the trouble IMO.