[Buildroot] [PATCH] linux-headers: Account for LINUX_OVERRIDE_SRCDIR
arnout at mind.be
Thu Jan 12 22:18:22 UTC 2017
On 12-01-17 18:19, Andrey Smirnov wrote:
> On Thu, Jan 12, 2017 at 12:17 AM, Arnout Vandecappelle <arnout at mind.be> wrote:
>> On 11-01-17 20:56, Andrey Smirnov wrote:
>>> Use the value of LINUX_OVERRIDE_SRCDIR as override directory when
>>> BR2_KERNEL_HEADERS_AS_KERNEL is specified.
>> Good catch! However, I'm not entirely convinced that we really want this.
>> A typical workflow would be to start from an upstream kernel, build a toolchain
>> and userspace with it, and then begin with hacking on the kernel to e.g. add a
>> driver. In that case it's rather unlikely that we really want to rebuild the
>> toolchain, because the headers don't change.
> I agree that typical "good day" workflow would be just as you
> described. However, IMHO, there's also a "bad day" variant of that
> workflow where "hacking on the kernel" implies "hacking on kernel
> headers"(*cough* custom ioctls *cough*), which is also followed up by
> a user-space tool that uses those changes and as such has
> 'linux-headers' as a dependency.
Certainly, but in that specific case you can easily set both
LINUX_OVERRIDE_SRCDIR and LINUX_HEADERS_OVERRIDE_SRCDIR.
>> That said, another possible scenario is that you don't configure any kernel
>> version but just use the OVERRIDE_SRCDIR to point to a local directory. Since we
>> removed the BR2_LINUX_KERNEL_CUSTOM_LOCAL, it's the only way to support that
>> scenario. In addition, the principle of least surprise would suggest that
>> BR2_KERNEL_HEADERS_AS_KERNEL would really work all the time, even if
>> LINUX_OVERRIDE_SRCDIR is set.
>> In short, I'm not really sure what is the best thing to do here. On thing though...
>>> Signed-off-by: Andrey Smirnov <andrew.smirnov at gmail.com>
>>> package/linux-headers/linux-headers.mk | 1 +
>>> 1 file changed, 1 insertion(+)
>>> diff --git a/package/linux-headers/linux-headers.mk b/package/linux-headers/linux-headers.mk
>>> index 0900778..197efb4 100644
>>> --- a/package/linux-headers/linux-headers.mk
>>> +++ b/package/linux-headers/linux-headers.mk
>>> @@ -10,6 +10,7 @@
>>> ifeq ($(BR2_KERNEL_HEADERS_AS_KERNEL),y)
>>> LINUX_HEADERS_VERSION = $(call qstrip,$(BR2_LINUX_KERNEL_VERSION))
>>> +LINUX_HEADERS_OVERRIDE_SRCDIR = $(LINUX_OVERRIDE_SRCDIR)
>> ... At least we should use ?= here so it is still possible to override each one
> I am more than happy to replace '=' with '?=', but I have to ask, why
> would you ever want to do that? Why would you configure your build
> system such that kernel headers are "Same as kernel being build" and
> then set you kernel source and kernel header source to point at
> different locations?
Because the canonical way to work is that your BR2_LINUX_KERNEL_CUSTOM_REPO_URL
points to your custom repository, and you just use LINUX_OVERRIDE_SRCDIR while
doing the kernel development. So the application developers only use the updated
kernel after it's explicitly been released. The kernel developer typically
doesn't want to rebuild the toolchain when he switches on LINUX_OVERRIDE_SRCDIR.
Arnout Vandecappelle arnout at mind be
Senior Embedded Software Architect +32-16-286500
G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
More information about the buildroot