On Mon, 2022-06-27 at 16:12 +0000, Ross Burton wrote: > Whilst I agree with this series in general (obviously…), some > commentary. > > First, the impact of this can’t be understated. If you enable dev- > pkgs on core-image-base, and extra 424 packages are installed > *ignoring any package with -dev in its name*. This includes > behaviour-changing packages such as the full-fat versions of > coreutils, util-linux, findutils, most of Python, a whole slew of > libraries. That is *not* good and this is a step in the right > direction. > > There are still some edge cases though, repeating the same test with > core-image-base installs these extra non-dev packages: > > alsa-topology-conf > libatopology2 > libavahi-client3 > libavahi-glib1 > libavahi-gobject0 > libc-malloc-debug0 > libc6-extra-nss > libc6-thread-db > libc6-utils > libfdisk1 > libform5 > libformw5 > libgcc1 > libmenu5 > libmenuw5 > libncurses5 > libncursesw5 > libnl-3-cli > libnl-idiag-3-200 > libnl-nf-3-200 > libnl-xfrm-3-200 > libnss-db2 > libpanel5 > libpanelw5 > libpcrecpp0 > libpcreposix0 > libsmartcols1 > libss2 > libstdc++6 > libtic5 > libticw5 > libx11-xcb1 > libxcb-composite0 > libxcb-damage0 > libxcb-dpms0 > libxcb-dri2-0 > libxcb-dri3-0 > libxcb-glx0 > libxcb-present0 > libxcb-randr0 > libxcb-record0 > libxcb-render0 > libxcb-res0 > libxcb-screensaver0 > libxcb-shape0 > libxcb-shm0 > libxcb-sync1 > libxcb-xf86dri0 > libxcb-xfixes0 > libxcb-xinerama0 > libxcb-xinput0 > libxcb-xkb1 > libxcb-xtest0 > libxcb-xv0 > libxcb-xvmc0 > > The bulk of these are likely from the -dev package depending on the > real package to satisfy dangling symlinks. This may actually be > desirable behaviour? Not sure I’d want a broken -dev package.
I did a test build and changing package_fixsymlinks() in package.bbclaass to use RRECOMMENDS instead of RDEPENDS does remove the packages listed above too. I'm torn on that, making that change would complete the picture and you'd still have the dependency there, just as a slightly weaker level and means dev-pkgs no longer radically changes the contents of the image which is probably a good thing? Cheers, Richard
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#167329): https://lists.openembedded.org/g/openembedded-core/message/167329 Mute This Topic: https://lists.openembedded.org/mt/91997856/21656 Group Owner: openembedded-core+ow...@lists.openembedded.org Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
