Redmine › SCM Creator (+Github)

My writing definitely needs some work...

No, they can’t! If users change the path (and this path differs from the one specified in scm.yml) they will get an error.

I may have overstated the problem The issue I actually encountered is that I can name my repos anything I want (inside scm.yml's path), and that name may collide with another project’s.

For example, given a path /var/lib/svn: if I create a project test, I can create a repo at /var/lib/svn/test. But I can create another repo for test at /var/lib/svn/test2. If later on, another user creates a project named test2, he can’t use /var/lib/svn/test2 because that’s already taken. From the second user’s perspective, that seems like a bug.

Enabling fully automatic repo creation would mean that name collisions like this just can’t happen. And it would be easier for the users, too; just pick a name for your new repo and you’re good to go. This is how it is with Gitorious, for example.

My ultimate goal is something like Google Code/Github/Gitorious, where the user shouldn’t have to bother at all with where on the server his repos should go. They all will be created for him, readily accessible with nice URLs, and all using Redmine for authentication. That may be a little outside this plugin’s original calling, but it’s so close...

Anyway why do you need that main if there should be only one repo per project?.. I would recommend to go with Force repository creation for all projects in this case. Or I did not get something...

The one repo per project and forced repo creation is my current workaround for the issue. Since the user never gets to muck around with the URL, there’s no problem. I intend to enable multiple repos, except that letting my users edit the URL (almost) at will is asking for trouble...

Not sure if file:///var/lib/svn/test/main will work with Redmine-based authentication... Seems it should be file:///var/lib/svn/test.main instead...

Sure, why not
As long as Redmine-based auth works, the repos are created automatically and there’s no name collisions, the actual path is not a problem

Update: fixed the quoting.

Ok, I got it.

My ultimate goal is something like Google Code/Github/Gitorious, where the user shouldn’t have to bother at all with where on the server his repos should go. They all will be created for him, readily accessible with nice URLs, and all using Redmine for authentication. That may be a little outside this plugin’s original calling, but it’s so close...

With auto_create=true you get this, don’t you?

My ultimate goal is something like Google Code/Github/Gitorious, where the user shouldn’t have to bother at all with where on the server his repos should go. They all will be created for him, readily accessible with nice URLs, and all using Redmine for authentication. That may be a little outside this plugin’s original calling, but it’s so close...

With auto_create=true you get this, don’t you?

I did try it, but I can’t recall the results (gotta get an external HD for my brain ), and forced creation and one repo per project worked well enough.

I’m at home, so I can’t verify this right now, but this issue will pop up whenever the user gets an editable URL field.

Yes, I understand.
Feature request accepted! Will do this.

Andriy Lesyuk wrote:

I guess this issue should be resolved along with #2053... Otherwise how would it be possible to add more repositories? By accepting .<number>?

I didn’t quite get #2053... With it implemented, wouldn’t renaming the identifier move the repo around in the server’s filesystem? And isn’t the URL immutable once created? If so, for my situation at least, providing an identifier should be mandatory and immutable once given, just like a project identifier, shouldn’t it? Or am I missing something?

I didn’t quite get #2053... With it implemented, wouldn’t renaming the identifier move the repo around in the server’s filesystem? And isn’t the URL immutable once created? If so, for my situation at least, providing an identifier should be mandatory and immutable once given, just like a project identifier, shouldn’t it? Or am I missing something?

Saying you have right now (when you create a repository): repository identifier set to “” (empty) and URL set to file:///var/lib/svn/test-project.2. URL is immutable (that is a user can’t modify it). When you change repostory identifier to additional the URL is automatically changed to file:///var/lib/svn/test-project.additional (it’s still immutable for users, but not for JavaScript). As soon as you click “Create new repository” both – Identifier and URL – become immutable.

Without this URL for additional repos will be set to something like file:///var/lib/svn/test-project.2... And there will be no way to change.

Hope this makes sense? Your thoughts?

Andriy Lesyuk wrote:

Saying you have right now (when you create a repository): repository identifier set to “” (empty) and URL set to file:///var/lib/svn/test-project.2. URL is immutable (that is a user can’t modify it). When you change repostory identifier to additional the URL is automatically changed to file:///var/lib/svn/test-project.additional (it’s still immutable for users, but not for JavaScript). As soon as you click “Create new repository” both – Identifier and URL – become immutable.

Without this URL for additional repos will be set to something like file:///var/lib/svn/test-project.2... And there will be no way to change.

Hope this makes sense? Your thoughts?

OK, I got it now. This solution works almost perfectly for me! It would be perfect along #2078

Update: mentioning #2078 as well.

Andriy Lesyuk wrote:

When you change repostory identifier to additional the URL is automatically changed to file:///var/lib/svn/test-project.additional (it’s still immutable for users, but not for JavaScript). As soon as you click “Create new repository” both – Identifier and URL – become immutable.

Just as a side note, URL editability should also be enforced in the server code (not just JavaScript), or malicious users would otherwise be able to create repositories at arbitrary locations in the system by forging the appropriate requests.

Hi,

Any news on this ? I would really appreciate this option for the users of the University I’m working on. It would be even better if the url field doesn’t appear and we only ask the user for the identifier. This way, the user is not bothered with an url including "/var/lib/svn” since he only needs an url like “http://redmine.example.com/svn/<generated_id>”

Best regards,

Ludovic Schoepps.