[Ksummit-2013-discuss] [ATTEND] static checking; COMPILE_TEST

[Mark Brown]

On Tue, Jul 16, 2013 at 10:01:44PM +0200, Jiri Slaby wrote:
> On 07/16/2013 09:56 PM, Steven Rostedt wrote:

> > If the current tool reports are being ignored or even rejected by
> > the maintainers, what makes you think that COMPILE_TEST will be
> > any different?

> No, this is a different story. This was about maintainers weren't
> willing to add dependencies to Kconfigs. Like there is FB_GEODE which
> actually depends on X86_32, but there was no such "depends on" in
> Kconfig and they didn't want me to add one. It is due to their "urge"
> to be able to compile the driver even on X86_64 .configs. So we came
> up with COMPILE_TEST and we now have "depends on COMPILE_TEST ||
> X86_32". They make COMPILE_TEST=y and can compile it whenever, normal
> users make it =n...

Right, the issue COMPILE_TEST helps with is that there's never been good
consensus on if we should enable builds where possible for compile
coverage (which is generally useful for people doing development) or
bother people with config options for devices they can never run (which
leads to wasted resources for distros and so on building and installing
drivers).  This is particularly prevelant for embedded stuff where SoC
vendors have their own IPs they don't license but from a code point of
view Linux's abstractions often mean the drivers will compile fine for
totally unrelated architectures.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.linuxfoundation.org/pipermail/ksummit-2013-discuss/attachments/20130716/52f40d4e/attachment.sig>