Re: Weird bitbake generation behavior


Richard Purdie
 

On Wed, 2021-08-18 at 06:23 +0000, Frans Meulenbroeks via lists.yoctoproject.org wrote:
Hi,

I share an sstate-cache with my fellow developers and I was assessing why sometimes things got rebuild even
though I did not expect this.
One of the things I discovered was that we had two versions of of glog/0.3.5-r0 in sstate.
The difference was caused run.do_configure where one user had this in run.do_configure

do_configure() {
    cmake_do_configure
    # remove WORKDIR info to improve reproducibility
    if [ -f  "/workdir/build-nano/tmp/work/aarch64-sorama-linux/glog/0.3.5-r0/build/config.h" ] ; then
        sed -i 's/'$(echo /workdir/build-nano/tmp/work/aarch64-sorama-linux/glog/0.3.5-r0 | sed
's_/_\\/_g')'/../g' /workdir/build-nano/tmp/work/aarch64-sorama-linux/glog/0.3.5-r0/build/config.h
    fi
}

whereas the other just had:

do_configure() {
    cmake_do_configure
}


The weird thing is that these two builds were about a day apart, they were build on the same system with as
far as I know the same metadata, the same distro, the same image etc etc
User settings should also be the same (we build under docker and I checked, we used the same version of ubuntu
in the container (18.04). (Actually the containers were generated from the same Docker file and docker inspect
tells me the images are identical).

Anyone an idea how this happens and where that extra snippet comes from? (I grepped for the string
"reproducibility" in the bitbake folder, but that did not help)
BTW we're using dunfell
Google shows:

https://patchwork.openembedded.org/patch/176924/

so I'd check the glog recipe.

https://git.openembedded.org/meta-openembedded/commit/?id=be95549f2ea5c59c6da6ace852b918cdba3c7822

https://git.openembedded.org/meta-openembedded/commit/?h=dunfell&id=a51c0d9b3a0d7dc5986ce46e7de65ef26cd06373

Cheers,

Richard

Join yocto@lists.yoctoproject.org to automatically receive all group messages.