oe-selftest currently has some checks to ensure the configuration doesn't include values that have an adverse affect on the tests. This provides a jarring user experience, but we also do not want to have oe-selftest modify the user's configuration files itself. The right way to ensure oe-selftest has the environment it requires is to use a separate build directory, just for use by oe-selftest, and write out the required configuration (whilst inheriting certain values such as DL_DIR and *MIRRORS* from the primary configuration).
And DISTRO, and MACHINE, and... How do we draw the line between "user configuration which we ignore for tests" and "user configuration that is being tested"?
perhaps by pointing to s oe-selftest config that makes it explicit which variables we are excluding and including?
Already implemented but is part of oe-selftest threaded runs, http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/?h=alimon/oe_selftest_threaded&id=97e3066889e16a79e95966d2009c747db974c6af
The thing here what will be the right configuration template to use?, the initial created by oe-init-build-env?
Yes, I think it makes sense for the configuration used by oe-selftest to be a combination of the TEMPLATECONF values and the whitelisted variables (DL_DIR, SSTATE_CACHE, *MIRRORS) from the initiating environment.
What if I want to run selftest for all of the qemu machines? Will MACHINE be whitelisted to?
(In reply to comment #3) > Already implemented but is part of oe-selftest threaded runs, > > http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/?h=alimon/ > oe_selftest_threaded&id=97e3066889e16a79e95966d2009c747db974c6af New commit id, http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/?h=alimon/oe_selftest_threaded&id=a3fdad106080bdd5addc5c2bff34534ffb6db4cd
proposed patch sent to the ML http://lists.openembedded.org/pipermail/openembedded-core/2017-October/143698.html
this work is in progress but unfortunately it is still giving some errors in the autobuilder [1]. Current work is at [2] [1] complete log is attached [2] http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/log/?h=lsandov1/oe-selftest-isolation
Created attachment 4142 [details] Complete AU log from an internal oe-selftest run see comment 9 for more information about this text.
the reason that this bug depends on bug 12401 is that the current approach to complete this enhancement basically starts metadata/scripts which is not git controlled, so it is the same scenario.
Last (failed) build: https://autobuilder.yocto.io/builders/nightly-oe-selftest/builds/666
Due to bug 12385 targeted for M2, moving this one also.
Second revision sent to the mailing list [1] [1] http://lists.openembedded.org/pipermail/openembedded-core/2017-December/145743.html
FWIW I'm now toying with the idea of http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/t222&id=f65664c23e8fbefc821919c730399c5e099842a5 which puts the parallelism into unittest itself...
Patches to parallelise oe-selftest merged for 2.6.