x
/
git-upstream
Code Issues Proposed changes

Wrap markdown documentation to 80 characters 

The markdown files where not wrapped at 80 characters making it
difficult to read in plain text and not conforming the normal
documentation standards on OpenStack.

Change-Id: I028a721866e4a62ee87730a5d3d8050c913f2867
This commit is contained in:
Tom Cammann 2014-09-30 21:47:03 +01:00
parent 8ff1af4bd9
commit 9ad1992036
3 changed files with 227 additions and 84 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

30
KNOWN-ISSUES.md
View File

@ -2,9 +2,18 @@
## Problem description
If previous local changes were based off of a point before the previous import merge was complete, and were not rebased to be applied after this point before being merged, the import command will miss including these when discarding previous obsolete history.
If previous local changes were based off of a point before the previous import
merge was complete, and were not rebased to be applied after this point before
being merged, the import command will miss including these when discarding
previous obsolete history.
Given a previous import like the following where previously changes B & C were replayed, and the import was merged as E. Where subsequently a change D which was proposed before the import, is merged to mainline, a subsequent run of the tool to import K will fail to include D along with B', C' & G' when discarding the assumed unnecessary merges E & F. While these should be discarded, the tool needs to check to make sure if there are dangling changes that should be included.
Given a previous import like the following where previously changes B & C were
replayed, and the import was merged as E. Where subsequently a change D which
was proposed before the import, is merged to mainline, a subsequent run of the
tool to import K will fail to include D along with B', C' & G' when discarding
the assumed unnecessary merges E & F. While these should be discarded, the tool
needs to check to make sure if there are dangling changes that should be
included.
D----
/ \
@ -26,7 +35,8 @@ The tool will find the following commit path to consider:
## Implementing a fix (WiP)
With respect to the previous diagrams, the correct behaviour is to find the following set:
With respect to the previous diagrams, the correct behaviour is to find the
following set:
D----
\
@ -36,13 +46,21 @@ With respect to the previous diagrams, the correct behaviour is to find the foll
/
G
This requires walking the list of initial found (2nd diagram) and examining each merge commit up to the previous import to see if there is a change that shares history with the other parent. i.e. look at F and see if D contains a merge-base in common with E, an thus include. Or locate E and find any differences between E & L that should be included.
This requires walking the list of initial found (2nd diagram) and examining
each merge commit up to the previous import to see if there is a change that
shares history with the other parent. i.e. look at F and see if D contains a
merge-base in common with E, an thus include. Or locate E and find any
differences between E & L that should be included.
## Workaround
Waiting for a proper fix to be implemented, a workaround for this issue would be to perform the import only into a branch (usually master) which has no "pending" commits in other branches to be merged that span across different imports.
Waiting for a proper fix to be implemented, a workaround for this issue would
be to perform the import only into a branch (usually master) which has no
"pending" commits in other branches to be merged that span across different
imports.
For gerrit this means not having pending reviews spanning across different git-upstream import command runs.
For gerrit this means not having pending reviews spanning across different
git-upstream import command runs.
This could be achieved in two different ways:
1) merging all pending reviews before an import;

147
README.md
View File

@ -1,59 +1,92 @@
# What is git-upstream?
git-upstream is an open source Python application that can be used to keep in sync with upstream open source projects, mainly OpenStack.
git-upstream is an open source Python application that can be used to keep in
sync with upstream open source projects, mainly OpenStack.
The main usecase for this tool is for people who are doing active contributions to repositories that are mirrors of OpenStack repositories, with the intention that most of those contributions will be submitted to review.openstack.org at some point.
If you are running a public cloud based on OpenStack, having local changes needed to use it in your environment, you can use git-upstream to stay up to date with the upstream master in a easier way (with respect to using raw git commands).
The main usecase for this tool is for people who are doing active contributions
to repositories that are mirrors of OpenStack repositories, with the intention
that most of those contributions will be submitted to review.openstack.org at
some point.
If you are running a public cloud based on OpenStack, having local changes
needed to use it in your environment, you can use git-upstream to stay up to
date with the upstream master in a easier way (with respect to using raw git
commands).
git-upstream provides new git subcommands to support rebasing of local-carried patches on top of upstream repositories. It provides commands to ease the use of git for who needs to integrate big upstream projects in their environment. The operations are performed using Git commands.
git-upstream provides new git subcommands to support rebasing of local-carried
patches on top of upstream repositories. It provides commands to ease the use
of git for who needs to integrate big upstream projects in their environment.
The operations are performed using Git commands.
**Note**: currently git-upstream can be used only for projects that are maintained with Gerrit as it relies on the presence of Change-IDs.
Nevertheless, the code is quite modular and can be extended to use any part of commit message (e.g., other headers).
**Note**: currently git-upstream can be used only for projects that are
maintained with Gerrit as it relies on the presence of Change-IDs.
Nevertheless, the code is quite modular and can be extended to use any part of
commit message (e.g., other headers).
Specifically, current git-upstream version supports the following features
* **Single upstream branch import**
Your repository is tracking an upstream project and has local changes applied and no other branch is merged in. This can also be applied to tracking upstream packaging branches: *e.g.*, ubuntu/master => ubuntu/saucy-proposed/nova + local packaging changes.
Your repository is tracking an upstream project and has local changes applied
and no other branch is merged in. This can also be applied to tracking
upstream packaging branches: *e.g.*, ubuntu/master =>
ubuntu/saucy-proposed/nova + local packaging changes.
* **Multi branch import (upstream branch + additional branches)**
In this case, your project tracks an upstream repository, merges in an arbitrary number of branches and applies local carried changes.
In this case, your project tracks an upstream repository, merges in an
arbitrary number of branches and applies local carried changes.
* **Re-reviewing**
Reviewing (w/ Gerrit) of all locally applied changes if desired.
git-upstream creates an import branch in a manner that allows it to be fully re-reviewed or merged into master and pushed.
Reviewing (w/ Gerrit) of all locally applied changes if desired. git-upstream
creates an import branch in a manner that allows it to be fully re-reviewed
or merged into master and pushed.
* **Detailed logging**
git-upstream can output to both console and log file simultaneously. Multiple levels and these are managed separately for log file and console output.
This means jobs run by Jenkins can save a detailed log file separately as an artefact while printing status information to the console if those running the jobs don’t wish to have the console spammed with the details.
git-upstream can output to both console and log file simultaneously. Multiple
levels and these are managed separately for log file and console output.
This means jobs run by Jenkins can save a detailed log file separately as an
artefact while printing status information to the console if those running
the jobs don’t wish to have the console spammed with the details.
* **Dropping of changes that appear upstream**
Compares Change-Id's of changes applied since previous import with those that have appeared on the upstream branch since the last import point.
Compares Change-Id's of changes applied since previous import with those that
have appeared on the upstream branch since the last import point.
* **Interactive mode**
Once the list of changes to be re-applied has been determined (and those to be dropped have been pruned), the tool can open an editor (controlled by your git editor settings) for users to review those changes to be made and allow them to perform further operations such as re-ordering, dropping of obsolete changes, squashing.
Once the list of changes to be re-applied has been determined (and those to
be dropped have been pruned), the tool can open an editor (controlled by your
git editor settings) for users to review those changes to be made and allow
them to perform further operations such as re-ordering, dropping of obsolete
changes, squashing.
* **Dropping local changes**
It’s always possible for local changes to be superseded by upstream changes, so when these are identified and marked as such, we should drop them.
It’s always possible for local changes to be superseded by upstream changes,
so when these are identified and marked as such, we should drop them.
This can also occur where a change was applied locally, modified when being upstreamed based on review feedback and the resulting differences were ported to the internal as well. While the original change will be automatically dropped, also useful to drop the additional ported changes automatically if possible, rather than have it cause conflicts.
This can also occur where a change was applied locally, modified when being
upstreamed based on review feedback and the resulting differences were ported
to the internal as well. While the original change will be automatically
dropped, also useful to drop the additional ported changes automatically if
possible, rather than have it cause conflicts.
# What git-upstream is not
The name of this tool includes the "git-" prefix because of the Git naming convention that a Git subcommand must have.
So, as git-review (usually invoked with "git review [...]"), this tool can be invoked using "git upstream [...]".
The name of this tool includes the "git-" prefix because of the Git naming
convention that a Git subcommand must have. So, as git-review (usually invoked
with "git review [...]"), this tool can be invoked using "git upstream [...]".
That said, and even if git-upstream currently uses Change-Ids, it is not strictly related to git-review.
In other words, git-review can (and most of the time will) be used without even knowing about git-upstream existence.
That said, and even if git-upstream currently uses Change-Ids, it is not
strictly related to git-review. In other words, git-review can (and most of the
time will) be used without even knowing about git-upstream existence.
# git-upstream installation
At the time of writing, there are two ways to install git-upstream: cloning its git repository or using pip.
At the time of writing, there are two ways to install git-upstream: cloning its
git repository or using pip.
## Installing from git repository
@ -81,7 +114,13 @@ Please see `USAGE.md`
### Description
Import code from specified upstream branch. Creates an import branch from the specified upstream branch, and optionally merges additional branches given as arguments. Current branch, unless overridden by the `--into` option, is used as the target branch from which a list of changes to apply onto the new import is constructed based on the the specified strategy. Once complete it will merge and replace the contents of the target branch with those from the import branch, unless `--no-merge` is specified.
Import code from specified upstream branch. Creates an import branch from the
specified upstream branch, and optionally merges additional branches given as
arguments. Current branch, unless overridden by the `--into` option, is used as
the target branch from which a list of changes to apply onto the new import is
constructed based on the the specified strategy. Once complete it will merge
and replace the contents of the target branch with those from the import
branch, unless `--no-merge` is specified.
### Usage
@ -124,7 +163,8 @@ optional arguments:
### Description
Mark a commit as dropped. Marked commits will be skipped during the upstream rebasing process.
Mark a commit as dropped. Marked commits will be skipped during the upstream
rebasing process.
See also the "git upstream import" command.
@ -148,20 +188,32 @@ optional arguments:
### Note
Commits will be marked with git notes in the namespace `refs/notes/upstream-merge`.
To list of commit id marked with a note, run `git notes --ref refs/notes/upstream-merge`.
To show a specific note run `git notes --ref refs/notes/upstream-merge show <marked commit sha1>`
Commits will be marked with git notes in the namespace
`refs/notes/upstream-merge`.
To list of commit id marked with a note, run `git notes --ref
refs/notes/upstream-merge`.
To show a specific note run `git notes --ref refs/notes/upstream-merge show
<marked commit sha1>`
As `drop` uses git notes to mark commits that have to be skipped during import, notes should be present on the cloned copy of your repository. Thus, if you are going to create notes on a system and perform the actual import on a different system, **notes must be present on the latter**.
You can push notes directly to git repository on the target system or push them in a different repository and then pull notes from your target system.
As `drop` uses git notes to mark commits that have to be skipped during import,
notes should be present on the cloned copy of your repository. Thus, if you are
going to create notes on a system and perform the actual import on a different
system, **notes must be present on the latter**.
You can push notes directly to git repository on the target system or push them
in a different repository and then pull notes from your target system.
## supersede
### Description
Mark a commit as superseded by a set of change-ids. Marked commits will be skipped during the upstream rebasing process **only if all the specified change-ids are present in `<upstream-branch>` during import**.
If you want to unconditionally drop a commit, use the `drop` command instead.
Mark a commit as superseded by a set of change-ids. Marked commits will be
skipped during the upstream rebasing process **only if all the specified
change-ids are present in `<upstream-branch>` during import**. If you want to
unconditionally drop a commit, use the `drop` command instead.
See also the "git upstream import" command.
@ -196,15 +248,25 @@ optional arguments:
### Note
*This command doesn't perform the actual drop*. Commits to be dropped during the next import, will be marked with git notes in the namespace `refs/notes/upstream-merge`.
There is no need to retain notes after an import dropped the correspondent commits, of course it doesn't harm keeping them either.
*This command doesn't perform the actual drop*. Commits to be dropped during
the next import, will be marked with git notes in the namespace
`refs/notes/upstream-merge`. There is no need to retain notes after an import
dropped the correspondent commits, of course it doesn't harm keeping them
either.
To list of commit id marked with a note, run `git notes --ref refs/notes/upstream-merge`.
To show a specific note run `git notes --ref refs/notes/upstream-merge show <marked commit sha1>`.
To list of commit id marked with a note, run `git notes --ref
refs/notes/upstream-merge`.
To show a specific note run `git notes --ref refs/notes/upstream-merge show
<marked commit sha1>`.
As `supersede` uses git notes to mark commits that have to be skipped during import, notes should be present on the cloned copy of your repository. Thus, if you are going to create notes on a system and perform the actual import on a different system, **notes must be present on the latter**.
You can push notes directly to git repository on the target system or push them in a different repository and then pull notes from your target system.
As `supersede` uses git notes to mark commits that have to be skipped during
import, notes should be present on the cloned copy of your repository. Thus, if
you are going to create notes on a system and perform the actual import on a
different system, **notes must be present on the latter**. You can push notes
directly to git repository on the target system or push them in a different
repository and then pull notes from your target system.
# Known issues
@ -215,12 +277,17 @@ git-upstream was written by Darragh Bailey <dbailey@hp.com>.
# Acknowledgements
Thanks to *Aleksander Korzynski* and *Stanisław Pitucha* for taking the original design spec and some basic manual steps and experimenting with initial implementations.
Thanks to *Aleksander Korzynski* and *Stanisław Pitucha* for taking the
original design spec and some basic manual steps and experimenting with initial
implementations.
To *Davide Guerri*, for picking up a rough python tool and turning it into something that was actually usable.
To *Davide Guerri*, for picking up a rough python tool and turning it into
something that was actually usable.
Also to *Jon Paul Sullivan* and *Monty Taylor* to listening and providing a sounding board for different approaches.
Also to *Jon Paul Sullivan* and *Monty Taylor* to listening and providing a
sounding board for different approaches.
And finally to *Coleman Corrigan* among numerous others who acted as willing guinea pigs for the original manual approach.
And finally to *Coleman Corrigan* among numerous others who acted as willing
guinea pigs for the original manual approach.
Hope this eventually helped save you time and some hair.

134
USAGE.md
View File

@ -1,8 +1,12 @@
# Importing from upstream: using git-upstream
**Note**: this guide assumes that you are using a branch named *master* to maintain your new features or bug fixes that sit on top of the upstream code of some project (probably somewhat related to OpenStack).
**Note**: this guide assumes that you are using a branch named *master* to
maintain your new features or bug fixes that sit on top of the upstream code of
some project (probably somewhat related to OpenStack).
It is also assumed you are tracking releases, which is only one of the possible approaches to upstream tracking. Another approach would be tracking the master tip of a project. Of course even other strategies are possible.
It is also assumed you are tracking releases, which is only one of the possible
approaches to upstream tracking. Another approach would be tracking the master
tip of a project. Of course even other strategies are possible.
## Install git-upstream on a development workstation
@ -49,9 +53,11 @@ usage: git-upstream [--version] [-h] [-q | -v] <command> ...
## Initial import of an upstream project
To explain the usage of the git-upstream tool We are going to use a real-world (but trivial) example.
To explain the usage of the git-upstream tool We are going to use a real-world
(but trivial) example.
We will be also using environment variables: in case you are keen on doing copy-paste, you will only need to change them to suite your environment.
We will be also using environment variables: in case you are keen on doing
copy-paste, you will only need to change them to suite your environment.
```bash
export REPO_NAME="jenkins-job-builder"
@ -85,7 +91,9 @@ git fetch --all
4) Push refs
Push refs defined upstream to the `origin` remote (*i.e.*, the internal copy of the repository with local patches) using the string `upstream` as prefix, also pushing tags.
Push refs defined upstream to the `origin` remote (*i.e.*, the internal copy
of the repository with local patches) using the string `upstream` as prefix,
also pushing tags.
```bash
git for-each-ref refs/remotes/upstream --format "%(refname:short)" | \
@ -93,7 +101,8 @@ git for-each-ref refs/remotes/upstream --format "%(refname:short)" | \
xargs git push --tags origin
```
You may want to repeat the last two commands before starting any new feature development or a bug fix.
You may want to repeat the last two commands before starting any new feature
development or a bug fix.
5) Check-out the first import commit (*e.g.*, tag or SHA1)
@ -109,9 +118,11 @@ git checkout -b import/$FIRST_IMPORT_REF $FIRST_IMPORT_REF
git checkout -b master
```
Now the tips of master, `$FIRST_IMPORT_REF` and `import/$FIRST_IMPORT_REF` should be pointing to the same commit.
Now the tips of master, `$FIRST_IMPORT_REF` and `import/$FIRST_IMPORT_REF`
should be pointing to the same commit.
Push local master branch to the remote origin, and make `origin master` the default when pushing commits.
Push local master branch to the remote origin, and make `origin master` the
default when pushing commits.
```bash
git push -u origin master
@ -120,7 +131,8 @@ git push -u origin master
## Writing your patches/features
Now start to develop new feature or fix bugs on master, as usual.
For this example, we are going to change the .gitreview file in order to use a local Gerrit server.
For this example, we are going to change the .gitreview file in order to use a
local Gerrit server.
echo "diff --git a/.gitreview b/.gitreview
index bb7e85c..f98d928 100644
@ -134,7 +146,8 @@ index bb7e85c..f98d928 100644
-project=openstack-infra/jenkins-job-builder.git
+project=my-projects/jenkins-job-builder.git" | patch -p1
Don’t forget to commit and push (after this step, you may want to use git review as usual)
Don’t forget to commit and push (after this step, you may want to use git
review as usual)
```bash
git commit -a -m "Set .gitignore content to use internal gating infra"
@ -145,12 +158,19 @@ Our master (local and remote) tip should be now pointing to the last commit.
## Importing single patches from upstream
Before implementing any feature or fixing any bug (in short, before reinventing the wheel), check if someone has already implemented the required code upstream.
Before implementing any feature or fixing any bug (in short, before reinventing
the wheel), check if someone has already implemented the required code
upstream.
If not, try not to develop code only for your specific needs, be ambitious and try to develop something that could be useful for the whole community. This way you can propose your patch upstream and save yourself a lot of trouble which arise when there are many local changes to carry on the tip of upstream releases.
If not, try not to develop code only for your specific needs, be ambitious and
try to develop something that could be useful for the whole community. This way
you can propose your patch upstream and save yourself a lot of trouble which
arise when there are many local changes to carry on the tip of upstream
releases.
In this example, we tried to use our code and we found out that the job filtering isn’t working!
Fortunately, Antoine Musso has already fixed this bug, as we can see in the upstream repo.
In this example, we tried to use our code and we found out that the job
filtering isn’t working! Fortunately, Antoine Musso has already fixed this bug,
as we can see in the upstream repo.
```bash
git show --summary 2eca0d11669b55d4ab02ba609a15aa242fd80d14
@ -175,7 +195,8 @@ Date: Mon Jun 24 14:36:52 2013 +0200
Tested-by: Jenkins
```
We are also interested in the following commit, which adds the Environment File Plugin (finally!).
We are also interested in the following commit, which adds the Environment File
Plugin (finally!).
```bash
~# git show --summary bf4524fae25c11640ef839aa422ac81bd926ca20
@ -197,7 +218,8 @@ Date: Mon Jul 1 11:21:24 2013 -0700
Tested-by: Jenkins
```
Import those changes simply cherry-picking the two commits. Don’t forget to push (review!) your changes.
Import those changes simply cherry-picking the two commits. Don’t forget to
push (review!) your changes.
```bash
git cherry-pick 2eca0d11669b55d4ab02ba609a15aa242fd80d14
@ -216,9 +238,13 @@ git for-each-ref refs/remotes/upstream --format "%(refname:short)" | \
xargs git push --tags origin
```
A lot of work has been done upstream and we need to rebase our master onto the upstream master branch. In this process we must skip all the commits we already cherry-picked some days ago, of course.
A lot of work has been done upstream and we need to rebase our master onto the
upstream master branch. In this process we must skip all the commits we already
cherry-picked some days ago, of course.
**Note**: the rebasing for this example is trivial but it is just to break the ice. Later in this guide we will address rebasing conflicts that can occur in the real world.
**Note**: the rebasing for this example is trivial but it is just to break the
ice. Later in this guide we will address rebasing conflicts that can occur in
the real world.
Create a new local branch with the new release tag as a starting point
@ -227,7 +253,8 @@ git branch import/0.6.0 0.6.0
```
## Running git-upstream
Finally, it is time to run git-upstream! Before doing so make sure the current branch is master
Finally, it is time to run git-upstream! Before doing so make sure the current
branch is master
```bash
git checkout master
@ -251,7 +278,10 @@ import branch: 'import/0.6.0'
What has just happened?
git-upstream has created a new branch named `import/0.6.0-base` which tip is set to the commit pointed by the release tag `0.6.0`, and has rebased all changes present in our local master which were not already present in the upstream new release (`import/0.6.0-base`) onto `import/0.6.0-base`.
git-upstream has created a new branch named `import/0.6.0-base` which tip is
set to the commit pointed by the release tag `0.6.0`, and has rebased all
changes present in our local master which were not already present in the
upstream new release (`import/0.6.0-base`) onto `import/0.6.0-base`.
You can see that running the following command
@ -259,10 +289,13 @@ You can see that running the following command
git log --graph --oneline --all --decorate
```
For this trivial example, the only commit not present in the upstream release was about the customisation of the .gitreview file.
For this trivial example, the only commit not present in the upstream release
was about the customisation of the .gitreview file.
The default strategy git-upstream uses to find duplicate entries is the comparison of Change-id entries in commit messages.
Of course, it’s not possible to compare directly the SHA1 for a commit because the cherry-picking changes the information used for SHA1 calculation
The default strategy git-upstream uses to find duplicate entries is the
comparison of Change-id entries in commit messages. Of course, it’s not
possible to compare directly the SHA1 for a commit because the cherry-picking
changes the information used for SHA1 calculation
---
**Note**: A git commit SHA1 is generated from the following information:
@ -275,12 +308,23 @@ Of course, it’s not possible to compare directly the SHA1 for a commit because
---
The local branch `import/0.6.0` now contains our local changes rebased onto the new upstream release. git-upstream has also merged this branch with the local master branch (with "ours" strategy) to allow the normal workflow (committing/merging to master for review).
The local branch `import/0.6.0` now contains our local changes rebased onto the
new upstream release. git-upstream has also merged this branch with the local
master branch (with "ours" strategy) to allow the normal workflow
(committing/merging to master for review).
**Note**: The "final" merging step is not mandatory. Of course you can keep a separate branch for each new import. On one hand this strategy allows a "cleaner" history as you will always have your local changes rebased on top of the exact copy of the upstream repository. On the other hand you will be creating a new branch every time you want to import upstream code.
You can customise the name of the import branch using the `--import-branch <branch name>` option.
**Note**: The "final" merging step is not mandatory. Of course you can keep a
separate branch for each new import. On one hand this strategy allows a
"cleaner" history as you will always have your local changes rebased on top of
the exact copy of the upstream repository. On the other hand you will be
creating a new branch every time you want to import upstream code.
You can customise the name of the import branch using the `--import-branch
<branch name>` option.
In principle, you could also replace your master branch (history) with the new import branch created by git-upstream... Unfortunately there is no way to do this without requiring ad-hoc intervention on cloned copies of the repository (aka do-not-do-that(TM))
In principle, you could also replace your master branch (history) with the new
import branch created by git-upstream... Unfortunately there is no way to do
this without requiring ad-hoc intervention on cloned copies of the repository
(aka do-not-do-that(TM))
To disable automatic merging, just use the `--no-merge` flag
@ -290,9 +334,12 @@ To disable automatic merging, just use the `--no-merge` flag
# Handling conflicts
Of course n the real world things are much more complicated. From time to time, during import, you will get rebasing conflict (for instance due to changes from both local and upstream repository to the same piece of code).
Of course n the real world things are much more complicated. From time to time,
during import, you will get rebasing conflict (for instance due to changes from
both local and upstream repository to the same piece of code).
In case of rebasing conflict, git-upstream will stop allowing the user to fix the conflict.
In case of rebasing conflict, git-upstream will stop allowing the user to fix
the conflict.
```bash
~# git-upstream import import/0.5.0 --into master
@ -309,7 +356,8 @@ Could not apply f9b4fca... Fixup for openstack review
Import cancelled
```
Let's find out why git-upstream failed and let's try to continue the rebasing manually.
Let's find out why git-upstream failed and let's try to continue the rebasing
manually.
```bash
~# git status
@ -337,10 +385,12 @@ Depending on the type of conflict, you will could:
* edit conflicting code
Change conflicting code in order to accomodate local changes to the new upstream code.
Change conflicting code in order to accommodate local changes to the new
upstream code.
You can later resume rebasing process issuing `git rebase --continue`
Currently git-upstream can't resume the rebasing process. So, if needed, the final "merging" steps have to be performed manually:
Currently git-upstream can't resume the rebasing process. So, if needed, the
final "merging" steps have to be performed manually:
```bash
git merge -s ours --no-commit <import-xxxx>
@ -358,17 +408,20 @@ Committing merge commit
git commit --no-edit
```
**Note**: git-upstream performs exactly those steps in order to replace the content of `master` branch with the import branch preserving the history.
**Note**: git-upstream performs exactly those steps in order to replace the
content of `master` branch with the import branch preserving the history.
# Integration with Gerrit
You may want to use review with Gerrit the output of git-upstream, in order to perform tests, gating, etc.
You may want to use review with Gerrit the output of git-upstream, in order to
perform tests, gating, etc.
You have 2 options for doing that:
## Re-review every new commit
In this case we want to review every new commit (wrt the last import). In order to do so, use the `--no-merge` flag of git-upstream import commend, and:
In this case we want to review every new commit (since the last import). In
order to do so, use the `--no-merge` flag of git-upstream import command, and:
```bash
git checkout import-xxxxx
@ -376,11 +429,14 @@ git push gerrit import-xxxxx-base:import-xxxxx
git review import-xxxxx
```
If there is more than one new commit, git-review will ask to confirm the submission of multiple changes.
If there is more than one new commit, git-review will ask to confirm the
submission of multiple changes.
## Re-review only the final merge commit
This would be possible by using the `--import-branch` option of import command and **pushing directly** (*i.e.*: bypassing Gerrit) the new branch to the local repo. For instance:
This would be possible by using the `--import-branch` option of import command
and **pushing directly** (*i.e.*: bypassing Gerrit) the new branch to the local
repo. For instance:
```bash
TIMESTAMP=$(date +"%Y%m%d%H%M%s")
@ -394,7 +450,9 @@ Then, create a valid `Change-Id` for the merge commit
git commit --amend -C HEAD --no-edit
```
Locally, git-review will still complain about the presence of N+M commits which would be committed BUT on the remote side all those commits will be recognised as already present in one of the two branch involved in the merge.
Locally, git-review will still complain about the presence of N+M commits which
would be committed BUT on the remote side all those commits will be recognised
as already present in one of the two branch involved in the merge.
```bash
git review -R -y master