Update all ebuilds to point to a branch or tag.

Update all ebuilds to point to a branch or tag.

Mark all ebuilds in chromiumos-overlay and chromeos-overlay as stable and
point them at the current version we have checked out.

This is useful in case we want to verify that all ebuilds are pointing at the
right commit hash when we are doing a build on a branch. It is intended to be
used prior to tagging and can be used instead of a preflight queue. It is an
alternative to cros_mark_all_as_stable that does not require a chroot and
does not increase the revision number on ebuilds.

Unlike cros_mark_all_as_stable, which only updates the latest stable ebuild
for a given board, this script updates all stable ebuilds for all boards,
including unused stable ebuilds. We do this for the sake of simplicity
because it's hard to know in advance which ebuilds are truly unused, and it
shouldn't hurt to update unused ebuilds.

This script does not support cros_mark_as_stable_blacklist. If there are any
packages in the blacklist, this script exits with an error message.

TEST=Ran this script to point power_manager at branch and tested that we
     built power_manager from the branch.
BUG=chromium-os:7218, chromium-os:6429
Change-Id: Ib2308806c5a4db979b1245a90407fd5b6353e4fd

Committed: http://chrome-svn/viewvc/chromeos?view=rev&revision=807d018

[sosa, 2010-10-08 22:03:10]

I would try not to mention specific roles that script might be of interest to as
devs etc may be interested in doing this as well for testing changes before they
commit to a branch.

http://codereview.chromium.org/3607014/diff/2002/4001
File cros_mark_branch_as_stable (right):

http://codereview.chromium.org/3607014/diff/2002/4001#newcode27
cros_mark_branch_as_stable:27: if [[ -z $FLAGS_branch ]]; then
You need quotes around $FLAGS_branch here to check the empty case.

[djmm, 2010-10-08 22:29:25]

LGTM

[David James, 2010-10-11 15:09:30]

PTAL. I've updated the script to also copy changes from the 9999 ebuild to the
stable ebuilds. This is quite helpful because developers usually expect that
their changes to the 9999 ebuild will be included, and doing this in the script
avoids a potentially error-prone manual step.

[sosa, 2010-10-11 17:05:08]

As a side note, maybe we should just remove the blacklist functionality from
cros_mark_all?  What do you think?

http://codereview.chromium.org/3607014/diff/18001/9002
File cros_mark_branch_as_stable (right):

http://codereview.chromium.org/3607014/diff/18001/9002#newcode55
cros_mark_branch_as_stable:55: unstable_ebuild=$(echo "$dir"/*-9999.ebuild |
is_workon | is_unstable)
Do we have 9999 ebuilds that are stable?  Do we really want to continue
supporting that or get rid of them if they appear?

http://codereview.chromium.org/3607014/diff/18001/9002#newcode57
cros_mark_branch_as_stable:57: sed -e '/^KEYWORDS="/s/~//g' -e
'/^CROS_WORKON_COMMIT=/d' \
My sed foo is somewhat weak.  Does this work if the 9999 ebuild doesn't already
have an empty CROS_WORKON_COMMIT id as most 9999 ebuilds don't have one.

http://codereview.chromium.org/3607014/diff/18001/9002#newcode68
cros_mark_branch_as_stable:68: echo Set branch to $FLAGS_branch
Would prefer to see quotes around this statement though I know it isn't strictly
necessary.

[David James, 2010-10-11 21:15:36]

Per feedback from Anush, I've updated to use commit hashes everywhere. This
makes the script a lot more complex.

This script is only needed as a temporary solution for aiding with branching.
Once all the features in here have been rolled into cros_mark_as_stable we can
kill this script.

http://codereview.chromium.org/3607014/diff/18001/9002
File cros_mark_branch_as_stable (right):

http://codereview.chromium.org/3607014/diff/18001/9002#newcode55
cros_mark_branch_as_stable:55: unstable_ebuild=$(echo "$dir"/*-9999.ebuild |
is_workon | is_unstable)
On 2010/10/11 17:05:09, sosa wrote:
> Do we have 9999 ebuilds that are stable?  Do we really want to continue
> supporting that or get rid of them if they appear?

We don't have any such ebuilds right now, but I've accidentally committed one
before. In the case where you accidentally mark a 9999 ebuild as stable, bad
things happen to users who are using the minilayout with cros-workon :(

We do have stable ebuilds (in the private overlay) which are cros-workon but
have no corresponding 9999 ebuild.

http://codereview.chromium.org/3607014/diff/18001/9002#newcode57
cros_mark_branch_as_stable:57: sed -e '/^KEYWORDS="/s/~//g' -e
'/^CROS_WORKON_COMMIT=/d' \
On 2010/10/11 17:05:09, sosa wrote:
> My sed foo is somewhat weak.  Does this work if the 9999 ebuild doesn't
already
> have an empty CROS_WORKON_COMMIT id as most 9999 ebuilds don't have one.
Yup. It just deletes any CROS_WORKON_COMMIT lines that are present, and then
adds a new CROS_WORKON_COMMIT after the EAPI line (same behavior as
cros_mark_as_stable)

http://codereview.chromium.org/3607014/diff/18001/9002#newcode68
cros_mark_branch_as_stable:68: echo Set branch to $FLAGS_branch
On 2010/10/11 17:05:09, sosa wrote:
> Would prefer to see quotes around this statement though I know it isn't
strictly
> necessary.

Done.

[sosa, 2010-10-11 21:21:01]

LGTM

[djmm, 2010-10-11 22:02:50]

LGTM

http://codereview.chromium.org/3607014/diff/25001/26001
File cros_mark_branch_as_stable (right):

http://codereview.chromium.org/3607014/diff/25001/26001#newcode49
cros_mark_branch_as_stable:49: eval $(grep -E '^CROS_WORKON' $ebuild)
Bash's elegant simplicity. :)

[anush, 2010-10-11 22:03:33]

LGTM

On 2010/10/11 22:02:50, djmm wrote:
> LGTM
> 
> http://codereview.chromium.org/3607014/diff/25001/26001
> File cros_mark_branch_as_stable (right):
> 
> http://codereview.chromium.org/3607014/diff/25001/26001#newcode49
> cros_mark_branch_as_stable:49: eval $(grep -E '^CROS_WORKON' $ebuild)
> Bash's elegant simplicity. :)