1. 03 Jun, 2019 1 commit
  2. 28 May, 2019 1 commit
  3. 13 May, 2019 4 commits
    • Harald Sitter's avatar
      statefile goes into wok not raid · cca3a597
      Harald Sitter authored
    • Harald Sitter's avatar
      set the screenshot interval from 0.5 to 0.3 · 95bcd954
      Harald Sitter authored
      I have high hopes that this makes cursor blinking less of an issue
    • Harald Sitter's avatar
      repair post-install tests · 66a049d6
      Harald Sitter authored
      kind of anyway. with the new os-autoinst snapshotting is a bit of a mess
      and so is the raid/ directory. it's all in all much harder to get
      efficient post-install tests going because keep_hdd mode (which is what
      we use for post-install) no longer creates a new "profile" without also
      creating a new hard disk, but instead loads the qemu_state of a previous
      profile. IOW a post-install test would load the qemu drives of the
      install test it is derived from. that technically is lovely but for two
      things. a) snapshotting seem to result in tianocore not booting after
      the snapshot. no clue why but that breaks uefi. b) we don't actually
      want the neon.iso be required, but it would be because the cd drive is
      based on it and qemu wouldn't start if it can't find the backing iso file.
      as a somewhat reliable fix for this the raid recycling is changed a bit.
      - install tests now disable snaphots, this prevents uefi from getting
        messed up for unknown reasons to do with snapshotting. plus we don't
        use snapshots anyway, so there's no reason why it would be enabled.
      - the !installation test setup now also copies the pflash drives (we
        only test on uefi, so the pflash drives are always present and necessary)
      - the qemu_state.json is is now moved into the raid/ dir after the test
        and subsequently picked up by !installation tests and copied into their
      - the qemu_state.json is further mangled to entirely drop the cd drive and
        by extension the requirement on neon.iso
      this should make !installation tests work once again. it's a bit of a mess
      though, conceptually speaking this is really far removed from how an
      ordinary openqa install would do this :/
    • Harald Sitter's avatar
      set an arch · f9507970
      Harald Sitter authored
      we don't technically need this but there is some logic which relies on
      ARCH being set, so best save than sorry. lest they use ARCH in a more
      required manner moving forward
  4. 09 May, 2019 1 commit
  5. 08 May, 2019 7 commits
  6. 07 May, 2019 4 commits
  7. 06 May, 2019 5 commits
  8. 12 Apr, 2019 2 commits
  9. 11 Apr, 2019 2 commits
    • Harald Sitter's avatar
      typo-- · f29d16e6
      Harald Sitter authored
    • Harald Sitter's avatar
      fully port away from xenial · d85841ad
      Harald Sitter authored
      the tests were somewhat forgotten when killing off xenial as they all
      are forwards and backwards compatible. seeing as xenial is no more
      we can also drop special workarounds done for it though.
      - all ENV fallbacks are gone
      - all conditionals in the test WRT bionic vs. xenial are gone and bionic
        is now the unconditional default expectation
  10. 27 Mar, 2019 2 commits
  11. 27 Sep, 2018 1 commit
  12. 20 Sep, 2018 1 commit
    • Harald Sitter's avatar
      rejigger the ubiquity partitioning test · bb43fba1
      Harald Sitter authored
      while this still sucks because it is part of ubiquity test proper, this is
      no longer part of the standard installation test.
      instead the partitioning test only runs when the correct var is set but it
      is still part of the regular installer code.
      additionally the BOOTFROM setting is now also conditional to the var, as
      unfotunately this behaved as I feared and prevents disk booting in
      first_start, so we cannot use this var for regular installation tests where
      we need to boot from disk
  13. 19 Sep, 2018 1 commit
    • Harald Sitter's avatar
      explicitly boot from cdrom when installing · ba4c944f
      Harald Sitter authored
      we expect the cd to get 'ejected' before rebooting, if that is not the case
      we'll consider this a problem anyway. I hope this works as expected
      this change is meant to facilitate a power reset when installing via
      ubiquity twice. on the boot after reset we come up with the disk already
      even though we didn't eject the cd, so I am expecting qemu simply boots
      from the hard disk because it has data and is ordered before the cdrom, so
      if we force cdrom to be preferred we'd ideally always boot from cdrom in
      installation tests except for when we eject it (which we do for the final
  14. 10 Sep, 2018 1 commit
  15. 06 Sep, 2018 3 commits
  16. 05 Sep, 2018 1 commit
    • Harald Sitter's avatar
      encode the type in the vars · bc63238c
      Harald Sitter authored
      we should be moving away from environment if we want to use openqa webui
      as I do not think we can pass environment vars from
  17. 08 Aug, 2018 3 commits
    • Harald Sitter's avatar
      exclude more crap from archival · ceb54fe6
      Harald Sitter authored
      we rsync because originally we reused the clone. we don't do this anymore.
      it's less invasive to exclude more of the large stuff than it is to
      refactor this to only archive the raid tho
    • Harald Sitter's avatar
      archive into series specific directories · 33f77124
      Harald Sitter authored
      this allows QAing two series moving forward with xenial and bionic it
      has become very apparent that we need proper QA running on two series when
      moving from one to the other and for a long time (as long as it takes
      to actually finish the move)
      (bionic builds are still not archiving for space concerns, will probably
       be enabled in limited capacity)
    • Bhushan Shah's avatar
      New location for the bionic ISO · 016422cc
      Bhushan Shah authored
      bionic-preview will be now public.