Here's a status of the current state of upstream repo branches "master"
and "1.9.x". But before getting to it, I would like to state that I
really had no pleasure in doing these reverts. I really do mean it. I
wish we had never ended up in this situation (and hopefully will never
again), but I do believe reverting this was the right decision.
1.9.x branch:
- Before starting the revert operation, the latest commit on this branch
was 7df9120ebc1f9bee97a6a1a47f0a5fda986e4ab0 which was the trailing
whitespace commit.
- Reverted that commit by issuing a "git revert
7df9120ebc1f9bee97a6a1a47f0a5fda986e4ab0"
- Did a clean build (just compilation and jar generation, no test cases
were run locally)
- Finally to make sure I didn't introduce any unnecessary
changes/problems of my own in this efforts, I compared the latest commit
(which was the revert commit) against the
7df9120ebc1f9bee97a6a1a47f0a5fda986e4ab0 commit's parent on 1.9.x,
b2da27513c7806357bf146bde44b3cc469757122 using the following command:
git diff b2da27513c7806357bf146bde44b3cc469757122
9b1b8dbbc6e9aa98922da28683f3773f586811e5
This returned empty (which is a good thing)
Pushed this state to upstream 1.9.x branch.
master branch:
This was a bit more complicated than the 1.9.x branch, given merge
commits plus additional master only commits that had happened on this
branch between the trailing whitespace commit, plus the fact that the
trailing whitespace commit touched extremely large number of files.
- Before starting the revert on this branch, the latest commit on it was
4ce54bf3b6c521af9c8db3229df5cd8b3199a3b2.
- The trailing whitespace commit was done before that above commit and
was 2f64e0b51c295960cb15aa77c7c1f447b2518e14. There were some other
unrelated commits between these 2 commits.
- Reverted that commit by issuing:
git revert 2f64e0b51c295960cb15aa77c7c1f447b2518e14
Needed to sort out merge conflicts with the following files:
Unmerged paths:
(use "git reset HEAD <file>..." to unstage)
(use "git add <file>..." to mark resolution)
both modified: src/etc/testcases/taskdefs/cvspass.xml
both modified: src/etc/testcases/taskdefs/delete.xml
both modified: src/etc/testcases/taskdefs/jar.xml
both modified: src/etc/testcases/taskdefs/optional/pvcs.xml
both modified:
src/etc/testcases/taskdefs/optional/xml/endpiece-ns-no-location.xml
both modified: src/etc/testcases/taskdefs/optional/xml/endpiece.xml
both modified: src/etc/testcases/taskdefs/typedef.xml
both modified: src/tests/antunit/taskdefs/copy-test.xml
both modified: src/tests/antunit/taskdefs/get-test.xml
Manually fixed those conflicts and completed the revert.
- Did one round of local clean build to make sure things were fine.
- At this point, the revert was complete, but there remained one more
step to merge 1.9.x branch into master.
- Issued "git merge 1.9.x" (where 1.9.x was the branch which included
the fully reverted state)
- Merge conflicts had to be resolved in the following files:
Unmerged paths:
(use "git add/rm <file>..." as appropriate to mark resolution)
both modified: src/etc/poms/ant-javamail/pom.xml
both modified: src/etc/poms/ant-swing/pom.xml
both modified: src/etc/poms/ant-testutil/pom.xml
both modified: src/etc/testcases/filters/build.xml
both modified: src/etc/testcases/taskdefs/conditions/antversion.xml
both modified: src/etc/testcases/taskdefs/delete.xml
both modified: src/etc/testcases/taskdefs/java.xml
both modified: src/etc/testcases/taskdefs/optional/pvcs.xml
both modified: src/etc/testcases/taskdefs/optional/script.xml
both modified: src/tests/antunit/core/extension-point-test.xml
deleted by us: src/tests/antunit/taskdefs/apt-test.xml
both modified: src/tests/antunit/taskdefs/get-test.xml
both modified: src/tests/antunit/taskdefs/javac-test.xml
both modified: src/tests/antunit/taskdefs/tar-test.xml
both modified: src/tests/antunit/taskdefs/taskdef-antlib-test.xml
Manually resolved the conflicts in all of those and committed the merge
locally.
- Ran one more round of local build and it went fine (no tests were
executed locally).
- Finally to make sure that I hadn't messed up these revert + merge, I
did a compare of the latest master commit (which included the revert +
merge state) against the parent of the (reverted)
2f64e0b51c295960cb15aa77c7c1f447b2518e14 commit,
aad5b519563cfe3dab3034be9bd3e83c0fb508c0 using the command:
git diff aad5b519563cfe3dab3034be9bd3e83c0fb508c0
995b518abf60dd8cd52f9e94c4186bbf78513b96
(unlike the 1.9.x case), this command returned a diff which contained
the expected set of changes that had gone in between the trailing
whitespace, force updated merge commits on master. Reviewed this
comparison diff and it looked fine and correct to me. Pushed this state
to "master" branch of upstream repo.
At this point, these branches are in a state where they can be used for
regular development and other planned activities.
-Jaikiran
On 06/07/18 2:32 PM, Stefan Bodewig wrote:
On 2018-07-06, Jaikiran Pai wrote:
On 05/07/18 2:42 PM, Stefan Bodewig wrote:
On 2018-07-05, Jaikiran Pai wrote:
I personally believe that reviewing these meaningless changes is a
waste of time and energy. I'm in favour of rolling back the entire
commit set if that's what it takes.
+1
although reverting the commits in both branches and merging back the
1.9.x branch is likely going to end in an ugly merge that will need
review as well. Hopefullly a shorter one.
Yes, I agree.I plan to attempt this later tonight (around 8 hours
from now) if no one else gets to it before that. Just letting it know
here, so that we don't end up duplicating these efforts.
Thank you. I'm unlikely to get there before you do, would have tackled
it tomorrow. Stefan
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@ant.apache.org For additional
commands, e-mail: dev-h...@ant.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@ant.apache.org
For additional commands, e-mail: dev-h...@ant.apache.org
