cros_copy_upgrade_server can now start the devserver with either prefix.
This is to help the migration from Memento to ChromeOSUpdateEngine.
BUG=none
TEST=none (yet)
Review URL: http://codereview.chromium.org/3304013
BUG=chromium-os:3828
TEST=Boot with test image. Check that OEM partiotion is not mounted at /mnt/partner_partition
Review URL: http://codereview.chromium.org/3355019
Change-Id: Ib7c15010df7d7f30e49149c42034b26a13d69573
It turns out that my previous rebuild_blacklist code had a bug where
blacklisted packages would get rebuilt unnecessarily. This would result
in slow builds if icedtea was built.
BUG=chromium-os:6628
TEST=Mock case where prebuilts for icedtea are older than packages that
icedtea depends on.
Change-Id: I498b6499b78ab065109c1fe0e77969028ceff6cb
Review URL: http://codereview.chromium.org/3308027
Change-Id: I95bb4518bbfa2931407a90180686845fd9e928e5
BUG=chromium-os:6574
TEST=manually run test script and verified all 4 test cases passed
Review URL: http://codereview.chromium.org/3341024
Change-Id: I6b9333fcf1325dbe141ad387f52d85781de96819
BUG=chromium-os:6600
TEST=added 8 test cases for RunCommand (up from zero :-) Ran all tests and
verified all passed. No regression either
Review URL: http://codereview.chromium.org/3307022
- This is needed for devices which use a USB stick or MMC card as their
primary storage.
BUG=1150
TEST=Installed an image to a MMC card, and copied an image to a MMC card.
Review URL: http://codereview.chromium.org/3337020
Change-Id: Ieb7f7cf7b8f9c608a23c55176c49b6805e08d382
Some mod_for_test images don't have the files necessary to run
emerge. Install them.
BUG=none
TEST=Rerun package install
Review URL: http://codereview.chromium.org/3351016
Change-Id: Ib19986121a8988c6cae23527148a7b5d4a58663e
BUG=chromium-os:6577
TEST=(inside chroot) run 4 test cases from src/scripts/, using Ctrl+C to force abort (hence triggering "delete_prompt")
a.) (stdin tty) "./build_image --board=x86-generic < /dev/null", expected = no prompt and delete output dir; actual == expected
b.) (stdout tty) "./build_image --board=x86-generic > foo.txt", expected = no prompt and delete output dir; actual == expected
c.) (normal user case) "./build_image --board=x86-generic",
expected = prompt to delete output dir, if y, output dir is removed; actual == expected
d.) (normal user case) "./build_image --board=x86-generic", expected = prompt to delete output dir, if N, output dir is NOT removed; actual == expected
Review URL: http://codereview.chromium.org/3373003
For chromite, we often want to run shell commands directly instead of using the
array syntax currently accepted by RunCommand. This makes it easier to run
piped commands, for example.
BUG=chromium-os:6573
TEST=I'm using it in cros_build.
Change-Id: Ic846e257fec4e423b0253d06641eac232545f6f6
Review URL: http://codereview.chromium.org/3323019
Change-Id: I415f089bc253c1ff9919451cf62c11a27a0bcf6d
BUG=6507
TEST=run cros_mark_all_as_stable and see the three previously ignored packages being uprevved
Review URL: http://codereview.chromium.org/3372002
While testing parallel_emerge, I found more cases where it would fail
to crack cycles. The issue is that parallel_emerge simply cracks the first
cycle it finds that involves an edge, rather than all cycles that involve an
edge.
Finding all cycles that involve an edge without looping forever is a rather
difficult problem, so I've updated the algorithm to simply keep searching
until it runs out of cycles. Testing this, I haven't found any cases where
the performance of the cycle cracking is a problem. I updated the algorithm
to log its performance stats so we can track this.
TEST=emerge -ep portage
BUG=none
Change-Id: I1cb95ffe9d977b9f8d38626d2d6cdbb766c14669
Review URL: http://codereview.chromium.org/3340010
Historically, we used --usepkgonly in build_image to ensure that packages
were not accidentally built during this step. Recently, build_image was
updated to use --usepkg. The benefit of using --usepkg was that it would not
install obsolete binary packages to your image. It was counter-intuitive that
--usepkgonly would install packages to your image that are not actually
installed to your board root. The disadvantage, however, of --usepkg, was that
it sometimes tried to build packages afresh using the image as your build root,
and this failed with strange errors because you're not supposed to do that.
This change fixes build_image to give useful errors by switching back to
--usepkgonly. To fix the problem where --usepkgonly installs packages that are
not installed, we first run eclean-<board> to clean all unused packages from
the board root.
This change is a big improvement because a number of people have run into
strange issues with build_image due to this problem and have had trouble
debugging them.
This change was actually written by Sean Paul, and I am shepherding his change
through for him because he doesn't have a Chromium account yet.
BUG=chromium-os:6437
TEST=
1) build_packages && build_image should still work
2) build_packages && ./cros_workon start power_manager && build_image should
fail with an error that all versions of the power manager are masked. This
happens because you started working on the power manager, but did not build
the version of the power manager you were working on before installing it.
3) Assuming you are working on the power manager, build_packages && build_image
should succeed because your cros_workon'd version of the package is built.
4) ( ./cros_workon start power_manager && build_packages &&
./cros_workon stop power_manager && build_image ) should fail with the
same error message as case 2 for similar reasons.
5) ( ./cros_workon start power_manager && build_packages &&
./cros_workon stop power_manager && build_packages && build_image )
should work because you've built your new version of the stable package.
Change-Id: Ia3858c70997bc6f0ec0b6d1bfaede8d3272a0976
Review URL: http://codereview.chromium.org/3305010
BUG=chromium-os:6204
TEST=Ran mod_image_for_test.sh, mounted resulting image and verfied
disable_idle_suspend now shows up under /usr/share/power_manager
instead of /var/lib/power_manager
Change-Id: Ia34787b364cd5ffa199efc4a36b5d9d691a33c5a
Review URL: http://codereview.chromium.org/3326011
This change must accompany http://codereview.chromium.org/3330005/show
Change-Id: If57a726933e4a243e424bdc98ddd577081afe6e2
BUG=chromium-os:6345
TEST=manual
built an image, ran mod_image_for_test on it, then
split the image into partition blobs, mounted part_3 and
checked /etc/init/{openssh-server,iptables,saft}.conf
- they all have the #for_test string removed.
removed.
Review URL: http://codereview.chromium.org/3309006
USB/SDCard boot devices typicaly don't run software updater,
and won't have success ever set. We never want to try to
fall back to the empty B partition either. The workaround
for this is to set the successful bit regardless of actual
boot success.
bug=6395
Change-Id: I67c625804203b13be9a0c626c404fa38bafb5445
Review URL: http://codereview.chromium.org/3344008
BUG=6325
TEST=cros_workon start and stop for board and host and see it work
Change-Id: Ic0680a273391fdf93b6ebbe0e34497807f31f240
Review URL: http://codereview.chromium.org/3352002
I put portage when I meant to put portage_pkg, so it was running
code unconditionally that should have only ran when portage was
being upgraded. Oops!
We also should clear the dependency list, since we don't want to
list dependencies other than portage when we're installing portage.
TBR=zbehan
TEST=parallel_emerge
BUG=none
Change-Id: Ic2861d78913f89fc42cedd96ffb125e08287f662
Review URL: http://codereview.chromium.org/3318009
We had a build failure today because portage was upgraded, and
parallel_emerge did not restart after emerging portage. This resulted
in build breakage because parallel_emerge did not pick up all of the
changes (only some of them).
BUG=chromium-os:6394
TEST=Tested that parallel_emerge restarts after upgrading portage in root,
but not after upgrading portage in board.
Change-Id: I480dc2dc034afb5b9e0468ed6e4d16de608d3d32
Review URL: http://codereview.chromium.org/3304010
Change-Id: Id81fe4e7bd9cae7a51a45b6ea8a535c4ea9430f8
BUG=
TEST=Ran locally on top of previous build area.
Review URL: http://codereview.chromium.org/3337007
parallel_emerge now checks the use flags of installed packages against new
packages, and checks the use flags of binary packages against source packages,
so that, when the appropriate flags are used, we never install packages with
the wrong use flags.
I've also fix related bugs in rebuild_blacklist. Previously, rebuild_blacklist
was too aggressive -- it even blacklisted packages from being rebuilt if they
are explicit workon packages. This bug also causes problems for the use flag
detection code because packages in this list would not get rebuilt when
the use flags changed. So I fixed this too.
Also relatedly, I've fixed our code for checking whether --selective is
enabled to look at the --newuse and --reinstall=changed-use flags. These
flags actually enable --selective implicitly, so we need to honor that for
correct behavior.
BUG=chromium-os:6359
TEST=Made sure that packages get rebuilt / reinstalled when use flags change.
Change-Id: I2d945b552863410ec7b286c3ed55ae4a64336ed5
Review URL: http://codereview.chromium.org/3361002
Change-Id: Iddc1a4c1fe83877eff020f3e075ddba5e5594a57
BUG=
TEST=Tested with new dev server changes. Ran through a successful update.
Review URL: http://codereview.chromium.org/3358004
This also mounts the path to our ssh-agent socket (usually in /tmp) inside the
chroot so we can use our external agent.
TEST=None
BUG=None
Change-Id: I543e8b2527be9958c1158234f39ecc34fc9dd0df
Signed-Off-By: Elly Jones <ellyjones@chromium.org>
Signed-Off-By: sosa <sosa@chromium.org>
Review URL: http://codereview.chromium.org/3277006
cros_workon would clobber local edits to local_manifest in many cases
This is a quick fix to prevent it. The proper solution is to actually parse
local_manifest as an XML doc and modify the DOM. Not play tricks with grep.
BUG=chromium-os:6272
TEST=Ran cros_workon against missing local_manifest, auto-generated local_manifest, local_manifest with indented tags. local_manifest with multi-line tags and local_manifest with <remote tags.
Review URL: http://codereview.chromium.org/3227006
Change-Id: I008c11a43ac21336575445273453373645f96398
I'm lazy and sometimes copy and paste the output at the end of build_image
verbatim and accidentally include the newline at the end. This gives a
more useful output for that case.
Change-Id: I1a4abefa884a91cb75dfe2779c79b3ef4b60e807
BUG=none
TEST=./image_to_usb --from=../build/images/x86-generic/latest --to=/dev/sdX (and /dev/sde)
Review URL: http://codereview.chromium.org/3212012
BUG=4887
TEST=tested run_remote_tests and image_to_live still work. Ran script with machine with no crashes, as well as on a machine with a slew of powerd CHECK(false)'s.
Change-Id: Iffb6571d30d99d876f41972f92a7149a716035ee
Review URL: http://codereview.chromium.org/3276002
* Add dev-rec key to factory installer
* rename factory_install_shim output to be consistent with dev install shim
Change-Id: Ibf8f027edda67626af5c319b4daa164cb53ccfe7
BUG=4382
TEST=Build factory install, build dev install, build normal
Review URL: http://codereview.chromium.org/3286002
so that the original users' subversion access permissions are preserved inside the chroot
Change-Id: I486070b3c1a2dda169ae0a95982ba693574e001b
BUG=
TEST=
Review URL: http://codereview.chromium.org/3249008
Change-Id: I758c93596d5cbdd6b52b9acc82f4d6e19a326c9f
BUG=5518
TEST=Tested using all the options. All tests listed in the file pass.
Review URL: http://codereview.chromium.org/3107039
