On 2016-05-13 7:52 PM, Jate Sujjavanich wrote:
One of the patches had a single word subject line that was very common
in the kernel repository's commit logs, so the script decided it was the
first one in the series towards the end. After I made it less general,
the script fell through to just applying all of the patches.

Thanks for the update. Glad to hear it is working.

Bruce


Thanks

Jate


On Thu, May 12, 2016 at 9:05 PM, Bruce Ashfield
<[email protected] <mailto:[email protected]>> wrote:

    On 2016-05-12 7:23 PM, Jate Sujjavanich wrote:

        I am attempting to incorporate kernel-yocto.bbclass into my
        Freescale
        kernel. I had to make some changes in the early tasks, but I got
        configuration fragments working (which was my main goal).

        My kernel recipe's patches stopped working, however. They
        followed the
        more conventional base.bbclass patching logic. The script
        kgit-s2q skips
        over several of my patches with the autoresume logic.

        Do you have suggestions on what to try? I do have to tell you
        that I'm
        on dora.


    Hmmm. It is rare that the script detects the wrong fence post, but
    it can happen if the shortlogs are similar and the diffstats also
    happen to match.

    I don't suppose there's a set of layers that you can send me, or
    point me at where I can do a few test runs ?

    Outside of that, you can add the native sysroot to your path and
    run kgit-s2q by hand, and see exactly why it is detecting the
    wrong fence post patch (and hence skipping).

    That script is looking at the patches, and then checking the branch
    from the top down to see if a given patch is already on the branch.
    Once it finds a patch that isn't on the branch, it declares it the
    resume point and starts from there. So if you have patches with
    similar shorlogs/diffstats, see about changing them to be unique
    and the script shouldn't move through them, looking for the resume
    point.

    In Yocto 2.0+ I ended up dropping the autoresume logic, since with
    the matching logic I hinted at above, it was indeed possible to
    shoot through the middle and resume from the wrong point.

    Bruce


        Jate




--
_______________________________________________
linux-yocto mailing list
[email protected]
https://lists.yoctoproject.org/listinfo/linux-yocto

Reply via email to