On 07/01/2013 03:56 PM, Geert Uytterhoeven wrote:

On Mon, Jul 1, 2013 at 9:47 AM, Chen Gang [off-list ref] wrote:

quoted

quoted

On 07/01/2013 03:32 PM, Geert Uytterhoeven wrote:

quoted

quoted

On Mon, Jul 1, 2013 at 8:44 AM, Chen Gang [off-list ref] wrote:

quoted

quoted

quoted

quoted

On 07/01/2013 02:34 PM, Richard Weinberger wrote:

quoted

quoted

quoted

quoted

quoted

quoted

quoted

quoted

config COMPILE_TEST

quoted

quoted

quoted

quoted

        bool "Compile also drivers which will not load"
        default n
        help
          Some drivers can be compiled on a different platform than they are
          intended to be run on. Despite they cannot be loaded there (or even
          when they load they cannot be used due to missing HW support),
          developers still, opposing to distributors, might want to build such
          drivers to compile-test them.

          If you are a developer and want to build everything available, say Y
          here. If you are a user/distributor, say N here to exclude useless
          drivers to be distributed.

I guess, 'COMPILE_TEST' is really useful for UML.

No.

We have explained you more than once what the deal and why UML does not
have ioremap() and friends.

This time, it is not for UML, it is for 'COMPILE_TEST': the
'asm-generic' whether need consider it.

COMPILE_TEST is only meant to be used as a dependency in Kconfig files,
to make drivers that cannot work due to missing hardware invisible when
configuring your kernel.

Why "Despite they cannot be loaded there (or even when they load they
cannot be used due to missing HW support)" ?

quoted

quoted

There should be no #ifdefs in actual code that check for (the lack of)
CONFIG_COMPILE_TEST.

In ideal world, it should be no #ifdefs for all "CONFIG_*" in 'normal code'.

But "asm-generic" are not belong to 'normal code', it is one of the main
place to process "CONFIG_*".

If the platforms need support 'COMPILE_TEST', the asm-generic is the
main place for implementing it.

So I think our focus is on "whether the platforms guys need support
'COMPILE_TEST" for modules guys".

COMPILE_TEST is to be used as a dependency if the code would compile,
but would not run. Adding fixes in the code to make more code compile is not
the job of COMPILE_TEST.

But from the explanation of 'COMPILE_TEST' in its Kconig, it seems not
like what you said.

e.g."Despite they cannot be loaded there (or even when they load they
cannot be used due to missing HW support)"

Can we allow the module to 'COMPILE_TEST' under one platform which not
support the related HW.

If we follow your reasoning, we can add dummy variants of all hardware support
that lives in arch/*/include/asm to asm-generic, so everything would compile
for all platforms. But that's now what we want.

"If HW support, implement default one, else do nothing" is more like
"default work", not "generic work".

"asm-generic" has already added default implementation for all
platforms, it also checked whether the HW support or not, that means at
least more than 80% works has done.

So if we still call it "asm-generic", I recommend to implement dummy one.

Thanks.
-- 
Chen Gang