[yocto] [SDK] including kernel devsrc to the SDK failes

[Khem Raj]

On Mon, Oct 29, 2018 at 8:14 AM Bruce Ashfield
<bruce.ashfield at windriver.com> wrote:
>
> On 2018-10-29 11:12 AM, Khem Raj wrote:
> > On Mon, Oct 29, 2018 at 6:46 AM Bruce Ashfield
> > <bruce.ashfield at windriver.com> wrote:
> >>
> >>
> >>
> >> On 10/29/2018 9:10 AM, Khem Raj wrote:
> >>> On Fri, Oct 26, 2018 at 6:54 AM Bruce Ashfield
> >>> <bruce.ashfield at windriver.com> wrote:
> >>>>
> >>>>
> >>>>
> >>>> On 10/26/2018 4:28 AM, Måns Zigher wrote:
> >>>>> So it looks like the number of /bin/awk have increased so the patch will
> >>>>> fix one problem but there is more to fix. I have a "workaround" evenYup. That works too (as would a variable from the env), but we'll still
> > need a sed based patch in the short term.
> >>>>> though I am not so sure it is a workaround or not. The problem is fixed
> >>>>> adding to kernel-devsrc.bbappend
> >>>>>
> >>>>> do_install_append() {
> >>>>> # This fixes the rpm dependency failure on install of kernel-devsrc
> >>>>> depending on /bin/awk
> >>>>> cd ${D} || true
> >>>>> for i in $(grep -srI "!/bin/awk" | cut -d":" -f1); do
> >>>>> sed -i -e "s#!/bin/awk#!/usr/bin/env awk#g" $i
> >>>>> done
> >>>>> }
> >>>>>
> >>>>> The original solution can be found at
> >>>>>
> >>>>> https://gerrit.automotivelinux.org/gerrit/gitweb?p=AGL/meta-agl.git;a=blobdiff;f=meta-agl-bsp/meta-ti/recipes-kernel/linux/linux-ti-staging_%25.bbappend;h=43fa494a26c132b875c177acf0389163d5e34c06;hp=b2e5431400acef0a5372e5490ac4f464482d4b3b;hb=4bfc821810cdee47611c6d3e94d771971f51fa75;hpb=1cf6b17fd15fde569374c85d4df722784f4e9c67
> >>>>>
> >>>>> The best solution might be to add kernel patches but since the number of
> >>>>> /bin/awk have increased I think that this is fine in the kernel universe
> >>>>> but a problem in poky universe. So by adding it to kernel-devsrc it will
> >>>>> make sure that when the kernel get's tainted in the future it will not
> >>>>> break the rpm build. This is a "bug" only when using rpm ipk will not
> >>>>> detect it as a problem.
> >>>>
> >>>> I'm ok with this type of solution as well, since this is similar
> >>>> to what we've had to do with perf in the past (sed and modify versus
> >>>> patching). I can always patch and fix things in linux-yocto, but then
> >>>> another other kernel still suffers the issue.
> >>>
> >>>
> >>> can we fix it in upstream kernel to use something like
> >>> #!/usr/bin/env awk
> >>> may be ?
> >>
> >>
> >> Yup. That works too (as would a variable from the env), but we'll still
> >> need a sed based patch in the short term.
> >>
> >
> > short term is fine, but I think this should be fixed in upstream kernel.
>
> Also agreed. That was my primary suggestion as well, since even with
> that slow pace, we'd have it in all the kernel trees by the next
> Yocto release window.

thats reasonable