Potential problem: on `update`, a change to build/conf/local.conf is only detected on a source update Potential behavior: on `update`, change is detected even when no source updates occur I discovered this discrepancy while configuring Scarthgap, where local.conf may be populated using TEMPLATECONF. Setup: ``` test-bbs$ git clone https://git.openembedded.org/bitbake bitbake test-bbs$ git -C bitbake checkout a70c336790a9188aae67975fac6ca13579ad1d3e test-bbs$ cp bitbake/default-registry/configurations/oe-nodistro-master.conf.json oe-nodistro-custom.conf.json test-bbs$ sed -i '8{s/^\([[:space:]]*\).*/\1"rev": "a70c336790a9188aae67975fac6ca13579ad1d3e"/;}' oe-nodistro-custom.conf.json test-bbs$ sed -i '15{s/^\([[:space:]]*\).*/\1"rev": "be8cdcf13a658e9e81ff2f7b71d1c8c37a920ce7"/;}' oe-nodistro-custom.conf.json test-bbs$ sed -i '22{s/^\([[:space:]]*\).*/\1"rev": "d97cfeaa55ed42722998d3313c1857aa377f6881"/;}' oe-nodistro-custom.conf.json test-bbs$ ./bitbake/bin/bitbake-setup init --non-interactive --setup-dir-name oe-nodistro-custom oe-nodistro-custom.conf.json nodistro machine/qemuarm64 test-bbs$ . bitbake-builds/oe-nodistro-custom/build/init-build-env ``` Introduce change: ``` test-bbs$ echo "# Modification detection test" >> bitbake-builds/oe-nodistro-custom/build/conf/local.conf ``` local.conf change undetected: ``` test-bbs$ ./bitbake/bin/bitbake-setup update NOTE: Bitbake-setup is using [...]/test-bbs/bitbake-builds as top directory. NOTE: Reading configuration from local file [...]/test-bbs/oe-nodistro-custom.conf.json Configuration in [...]/test-bbs/bitbake-builds/oe-nodistro-custom has not changed. ``` local.conf change detected on a source change (e.g. BitBake): ``` test-bbs$ sed -i '8{s/^\([[:space:]]*\).*/\1"rev": "112bddd8fc684fbdd71139429251b127739f863b"/;}' oe-nodistro-custom.conf.json test-bbs$ ./bitbake/bin/bitbake-setup update [...] Upstream bitbake configuration changes were found: diff '--color=always' -uNr [...]/test-bbs/bitbake-builds/oe-nodistro-custom/build/conf-backup.20260324205715/local.conf [...]/test-bbs/bitbake-builds/oe-nodistro-custom/build/conf/local.conf --- [...]/test-bbs/bitbake-builds/oe-nodistro-custom/build/conf-backup.20260324205715/local.conf 2026-03-24 20:54:31.919838770 -0600 +++ [...]/test-bbs/bitbake-builds/oe-nodistro-custom/build/conf/local.conf 2026-03-24 20:57:15.544151275 -0600 @@ -5,4 +5,3 @@ # it is recommended to put them into a distro config, or to create # layer fragments from changes made here. # -# Modification detection test ``` This is also detected with a change to a different source (e.g. OE-C): ``` test-bbs$ sed -i '15{s/^\([[:space:]]*\).*/\1"rev": "aa441c5d7c306fb308cb7890a8ebe3cd218780b5"/;}' oe-nodistro-custom.conf.json test-bbs$ ./bitbake/bin/bitbake-setup update [...] Upstream bitbake configuration changes were found: [...] ```
But is this a real problem? Is there a use case? If you make modifications in build/conf, I don't think bitbake-setup should track what you do there and compare that against 'pristine' build/conf, if nothing has otherwise changed. That would be heavy handed: people modify build/conf all the time, and they don't need bitbake-setup telling them their local modifications caused the build to become 'out of date'. The build/conf comparison only kicks in when updates happened elsewhere (e.g. in config template, or layer revisions changed), because then build/conf may have changed as well, and it needs to be setup anew, and compared with the existing one. At that point there's no way to tell what was original, and what was added by the user. Bitbake-setup could potentially track that too (by putting a copy of the original build/conf into a local git under config/), but again, I do not see the use case clearly, but I do see increased code complexity that would result.
Thanks for the feedback, Alex! I do not have a sample use case where this is required. Related to the original finding, it would be convenient to be able to configure the default value of `--update-bb-conf` (per Setup). This is most relevant for sans-fragment builds (e.g. Scarthgap), where the "no" option would be handy.
Alex ?
Alright. Closing for lack of clear use case.