I have a bazaar repository on a shared server. I'd like to clean up the repo and set it up from scratch but maintain my history. I don't know how the repository was created initially (is there a way to find out?).
Can I take a branch and make that into my main shared repo?
Is this a viable process:
bzr init-repo --no-trees /home/bzr/myrepository
cd /home/bzr/myrepository
bzr init stable
cp /home/oldbzr/branch_taken_from_current_repo/* ./stable/
cp /home/oldbzr/branch_taken_from_current_repo/.bzr ./stable/
Thanks
A "branch" and a "repo" in Bazaar are totally separate concepts. You don't convert a branch into a repo. What you usually think of as a repo (in SVN or Git, for example) is actually a branch in Bazaar. What you want to do is create a new repo, then copy the old branch into the new repo.
You almost have it right, but you don't want to use "cp", you want to use "bzr branch". Note: You can usually use "cp" to copy branches except when you want Bazaar to move a branch into, out of, or across a repository -- then you need "bzr branch" to intelligently repack the history. So here is what you want to do:
bzr init-repo --no-trees /home/bzr/myrepository
cd /home/bzr/myrepository
bzr branch /home/oldbzr/branch_taken_from_current_repo stable
Note that I am not doing "bzr init" -- I don't want to create a new branch, just copy the old one. And I am not manually copying the old branch or its .bzr directory. If you copy the old branch's .bzr, it will not end up using the new repository. By doing a "bzr branch" it will go "oh hey, I am moving into a repository. Therefore, I will put all of my commit data into the shared repository, and just put a lightweight branch in 'stable'."
You can use just plain branch into your shared repo as mgiuca suggested, but you also can convert your standalone branch to use shared repository. For that your steps should be extended with bzr reconfigure call:
bzr init-repo --no-trees /home/bzr/myrepository
cd /home/bzr/myrepository
bzr init stable
cp /home/oldbzr/branch_taken_from_current_repo/* ./stable/
cp /home/oldbzr/branch_taken_from_current_repo/.bzr ./stable/
cd stable
bzr reconfigure --use-shared
So, if we omit cp then you can create a shared repository "around" your branch:
cd /path/to/my/branch
bzr info # you should see you branch is standalone,
# i.e. not using shared repo
bzr init-repo ../ # create shared repo in parent directory
bzr reconfigure --use-shared # convert standalone branch to repository branch
bzr info # now you should see your branch is using shared repo
Related
I use git submodule add <GitHubURL.git> to add projects as submodules to a main "parent" project repository.
Parent_Project_repo
- SubA_repo
- SubB_repo
- SubC_repo
- Sub_Problem_Child_repo
As I work in the submodules, I make commits within the Sub*_repo project (per usual, committing whichever files I have worked on). In the parent project, however, usually I am just making a single commit for all the submodules commits. This single commit of the submodule usually shows up like so in the Parent_Project_repo:
I have recently added a submodule which, from the parent repository, displays each individual file in the commit history instead of just accessing all the files/commits in one single "Subproject" commit object.
Of note:
this problematic submodule does not show up in the roster when I
invoke:
git submodule status
It is listed in the
Parent_Project_repo/.gitmodules
Parent_Project_repo/.git/config
file as a submodule & there is a corresponding
Parent_Project_repo/.git/modules/Sub_Problem_Child_repo/ folder.
How can I get the problem child into the roster and able to have all it's commits handled by the Parent_Project_repo as one object?
UPDATE:
The only difference I have discerned with the “problem child” submodule is that it doesn’t have a “historySha” key in the /.git/modules/config file, e.g.:
[atomGithub]
historySha = 1936e4c373c130860a8f92683b517dad713ec37
Also, these commands don't get the Problem child listed in the status:
$ git submodule update --init --recursive
$ git submodule update --recursive
$ git submodule init
...nor showing up on GitHub in the Parent project with a "# e78c392" which indicates I can double click the link and got to that repo instead of a copy inside the superproject (Parent).
If git submodule add <URL> was used to set up a git repository as a submodule within a "super-project" and the submodule is NOT committing within the super-project as a single object (i.e. if, when committing the submodule from within the Super project you are having to commit each individual file from the submodule), then follow these steps to reconfigure:
Save all work in the submodule.
Stage and commit all file changes.
Push all commits to GitHub.
Make a safety copy of the submodule folder if you are cautious
Delete the relevant submodule section from the super-projects .gitmodules file (i.e. remove the listing of the problem submodule).
Stage the .gitmodules changes git add .gitmodules
Delete the relevant sunmodule section from super-projects .git/config.
Run git rm --cached -rf PATH_TO_SUBMODULE where PATH_TO_SUBMODULE is the actual path to the submodule folder.
Run rm -rf .git/modules/PATH_TO_SUBMODULE.
Commit git commit -m "Removed submodule commit message".
Delete the now untracked submodule files: rm -rf PATH_TO_SUBMODULE
git submodule add <URL>
Test to see if a change (fwiw, modify two files) in the submodule is handled as a single object when committing in the super-project.
Success? Delete the safety copy of the submodule project.
I have a bazaar repository holding several branches. I recently removed one of the trees with bzr remove-tree path/to/branch followed by rm -r path/to/branch. Now if I understand correctly, the repository should still hold the branch history, since I never did bzr remove-branch. However, I can't figure out how to retrieve the branch to continue working on it. Can someone help?
You can use bzr heads --dead to see the heads (= most recent revisions) of deleted branches, including their global revision ids.
You can then do:
bzr branch -r REVISION_ID REPO_DIR BRANCH_DIR
Here, REVISION_ID is the id of the head that you want to restore, REPO_DIR is the directory that holds the repository, and BRANCH_DIR is the directory where you want the branch to be stored.
Edit: If the above doesn't work for some reason, you can also do:
bzr init BRANCH_DIR
cd BRANCH_DIR
bzr pull -r REVISION_ID .
BRANCH_DIR must be underneath the repository directory, of course.
(EDIT: Removed my question, leaving a link to the question that confused me, because there's some useful stuff in some of the comments here)
The question here confused me, but was actually answered fully. That question asked how to build a new repository with only a few of the branches of the original repository.
I'm not entirely sure from your post what you're looking for. I'm taking a stab at giving an answer; let me know if that's not what you need.
The following is a simple script to copy all branches from a source to a target repo. Note that it won't work if the branch directory names contain any whitespace. You will have to setup the target repository with bzr init-repo first.
#!/bin/sh
SOURCEREPO=$1
TARGETREPO=$2
if [ ! -d "$TARGETREPO/.bzr" ]; then
echo "$TARGETREPO is not a Bazaar repository; create one with bzr init-repo"
exit 1
fi
BRANCHES=`cd "$SOURCEREPO"; find * -name .bzr -exec dirname '{}' ';'`
for branch in $BRANCHES; do
mkdir -p "$TARGETREPO/$branch"
if [ ! -d "$TARGETREPO/$branch/.bzr" ]; then
echo "Cloning $branch"
bzr branch --use-existing-dir "$SOURCEREPO/$branch" "$TARGETREPO/$branch" \
&& bzr config -d "$TARGETREPO/$branch" --remove parent_location
else
echo "Existing branch in $TARGETREPO/$branch"
fi
done
Basically, it does a bzr branch sourcerepo/branchdir targetrepo/branchdir for all branches and then uses bzr config to get rid of the parent location setting for each copy of a branch (because that location will presumably soon disappear).
I'm not quite sure what you really want here. Adding a remote branch to your local repository and recreating a repository from another are completely different things. The linked question is about the second, and the answers are spot-on, so I will address what you originally intended to do.
First of all, it's confusing when you say "main trunk repository", because "trunk" typically means a "main branch", and branches and repositories are different concepts, so "trunk repository" doesn't really make sense.
It seems to me that you have a shared repository: a container directory for multiple branches. For the sake of clarity: you create a shared repository with the command bzr init-repo, and you create branches in it with bzr init (new empty branch), or with bzr branch (copy of another branch, local or not).
Let's say you have a shared repository organized like this:
└── /tmp/shared/repo
├── bugfix123
├── feature1
└── feature2
If you want to add remote branches of your teammates, you can do it with:
bzr branch url_of_remote_branch /tmp/shared/repo/somename
For example if your teammate has a branch at the url lp:~jack/proj/bugfix312 you could get it into your local repository with:
bzr branch lp:~jack/proj/bugfix312 /tmp/shared/repo/bugfix312
If later you decide you don't want to merge this branch, you can get rid of it with the commands:
bzr remove-branch /tmp/shared/repo/bugfix312
rm -fr /tmp/shared/repo/bugfix312
The first command only removes Bazaar's branch data, it keeps the directory intact. The second removes the working directory itself.
Let me know if you are looking for something else.
I want to learn to use git-svn. I have an svn local repository on my disk that I've checked out a while ago using something like this:
svn co http://myserver.com/mysvnrepo/trunk/ /mysvnrepo/
ls -a /mysvnrepo/
. .. .svn foo bar
This /mysvnrepo/ is HUGE, so I want to avoid re-downloading or copying the files at all costs.
I'm wondering if there's a way to git clone this local repo without downloading / copying anything (because it's already there).
I have this which seems to be what I'm looking for, but when I do that it doesn't quite give me what I expect.
cd /mysvnrepo/
git svn clone file://mysvnrepo/
ls /mysvnrepo/
. .. .git .svn foo bar
git status
# On branch master
#
# Initial commit
#
# Untracked files:
# (use "git add <file>..." to include in what will be committed)
#
# .svn/
# foo/
# bar/
I would expect git to detect foo and bar as "versioned and up-to-date".
According to the docs it seems that I need to use git svn init because git svn clone runs a fetch, which I certainly don't want. So I tried
git svn init --trunk=file:///mysvnrepo/
...but no luck.
I'm completely new to git, so my confusion is off-the-charts... am I doing something utterly wrong?
Thanks in advance
You cannot take a subversion snapshot and convert it into a git repository.
It sounds like you are trying to avoid a lengthy initialization of the git repository from svn: which ordinarily will try to ready your entire history. This can be done in another way, by limiting the fetch to recent history depending on how much history is relevant to you:
git svn clone -s -r 12334:HEAD https://svn.host.org/repo
Where 12334 is the earliest svn revision you are interested in and assuming that the repo is laid out in a standard svn way with branches and tags.
I'm trying a
git rebase --onto master myremote/master~21 myremote/master
to add the latest 21 commits from a remote repository on mine.
What git tells me is that there's a conflict — but how's that possible?
In my understanding it's just taking that 21 commits and applying them on top of my master. How can there be conflicts?
Thanks for help!
I'm doing that btw because somehow I messed up my git-svn repository (the remote), and there's 21 commits which I don't manage to commit to subversion. So I'm trying with a fresh git-svn clone, in which I'm adding those 21 commits.
There is conflict if:
master has commit that are not in myremote/master.
those commits include common files/changes with one the last 21 myremote/master commits.
If somehow the fresh git-svn clone has different SHA1 than the previous git-svn repo, then there is no close common ancestors, and the chances of conflicts are that much higher.
See "How to identify conflicting commits by hash during git rebase?" for illustrations of conflicts during a rebase.
One way to reset your local master to myremote/master would be to:
git checkout -b tmp myremote/master # local tmp branch from myremote/master HEAD.
git merge -s ours master # ignore completely master content
git checkout master
git merge tmp # fast-forward to tmp HEAD
If you hadn't made any changes in your local master before fetching myremote/master, this should work.