[OE-core] [PATCH] kernel: do verbose builds

Bruce Ashfield bruce.ashfield at gmail.com
Mon Sep 16 20:41:53 UTC 2019


On Mon, Sep 16, 2019 at 4:27 PM Ross Burton <ross.burton at intel.com> wrote:
>
> On 16/09/2019 14:18, Bruce Ashfield wrote:
> > On Mon, Sep 16, 2019 at 8:41 AM Ross Burton <ross.burton at intel.com> wrote:
> >>
> >> Pass V=1 to kernel builds so that the compile log is more useful if it all goes
> >> wrong.
> >>
> >> Signed-off-by: Ross Burton <ross.burton at intel.com>
> >> ---
> >>   meta/classes/kernel.bbclass | 2 +-
> >>   1 file changed, 1 insertion(+), 1 deletion(-)
> >>
> >> diff --git a/meta/classes/kernel.bbclass b/meta/classes/kernel.bbclass
> >> index ebcb79a5286..9d2e1e476d4 100644
> >> --- a/meta/classes/kernel.bbclass
> >> +++ b/meta/classes/kernel.bbclass
> >> @@ -324,7 +324,7 @@ kernel_do_compile() {
> >>          fi
> >>          cc_extra=$(get_cc_option)
> >>          for typeformake in ${KERNEL_IMAGETYPE_FOR_MAKE} ; do
> >> -               oe_runmake ${typeformake} CC="${KERNEL_CC} $cc_extra " LD="${KERNEL_LD}" ${KERNEL_EXTRA_ARGS} $use_alternate_initrd
> >> +               oe_runmake ${typeformake} CC="${KERNEL_CC} $cc_extra " LD="${KERNEL_LD}" ${KERNEL_EXTRA_ARGS} V=1 $use_alternate_initrd
> >
> > Personally, I wouldn't want this on all the time. Most of what V=1
> > provides, I never use for debugging and it just clutters the logs.
> >
> > Can't this be triggered off a packageconfig-like setting for those
> > that want it ?
>
> If you object then I guess it's easy enough to add it to
> KERNEL_EXTRA_ARGS or something when required.  I only did this because I
> was chasing that build paths thing and needed the full build log, and
> typically we enable the full build log so it's there if required.

'object' is a strong word :D I just know that for the most part, I'm
not looking at CFLAGS or other Make/path issues when I'm looking at
the kernel build for issues (and that could be different for other
folks), and that I've found I end up doing custom instrumentation to
debug Make issues, versus getting anything really useful out of V=1
(it's a bit chatty). That and I've never done production builds with
V=1 (even if it won't impact the output).

It would make the logs a bit bigger, but I can't see it having a
performance impact .. so really, it isn't a big deal. But it sounded
like a knob that we could add, since clearly there are two different
points of view and modes that people may use.

Cheers,

Bruce

>
> Ross



-- 
- Thou shalt not follow the NULL pointer, for chaos and madness await
thee at its end
- "Use the force Harry" - Gandalf, Star Trek II


More information about the Openembedded-core mailing list