cm fast-import doesn't import owner(author) of the changeset properly

[b.kot]

Hi.

 

After I execute the command

 

cm fi myrepo svn-git-no-data.dump

 

"myrepo" is filled with changesets that are marked as created by "b.kot". The problem is that some changesets were made by "j.smith" and this information was lost during an import.

Below is the snippet from the import file created by: git fast-export -M --all

Here the same changeset (same comments "Próba formatowania[...]" and timestamp "2010-11-19 11:32:01") with the wrong author(should be j.smith - not b.kot)

I don't know if this is relevant, but the git repository was made from svn repository using svn-all-fast-export
The version of plastic is 4.1.10.391.
Could someone tell me what I have been doing wrong? I would appreciate any suggestions.

So far I have tried:

- granting all permissions for all users on myrepo

- various git fast-export switches configurations: -C -M --all|master

- activating license for user "j.smith"

- restaring PlasticSCM server

- creating repository before importing

 

 

[manu]

Hi!

 

Sometime I've seen that behavior during the import. I did the fast-import again on a new repository and all the authors were correctly migrated.

 

Can you try the import again and tell us if it works fine?

 

Best regards.

[b.kot]

Hi.

Unfortunately the 13th attempt brought the same results.
The thing is that  few days ago I managed to import data from the same repository without any major problems (The minor problem was that import somehow missed "master" branch although it was included in the export file - feature branches were imported smoothly). Then I used the same scripts on updated repository and bang... no authors.

 

I am considering now two options:
a) setting authors directly in a database repository (parse import file -> create map <comment, timestamp>: <commiter id>, identify row in plastics DB by <comment, timestamp>  -> update author according to map
b ) making backup of repos, reinstalling the plastics server and trying import one more time. The problem with this options is that we have already started working with Plastic. Preserving current changes are more import than preserving history.

However to perform first scenario I would need few pointers: where (in which table) should I search for the commits (changsets) metadata? where can I found users ID? Can you provide me this information?

 

 

Best regards.

[manu]

I'm afraid that a re-installation of Plastic is not going to help. The databases are going to be the same, only the binaries are replaced.

 

The DB process you mention is a bad idea, hard to do it and dangerous. I would prefer to set a new Plastic SCM server in a tiny linux box machine and try the import there, in a clean environment.

 

You also have the option to send us the package and we'll test it locally in order to find if there's any issue. You can use the "--no-data" parameter in order to generate a codeless package.

[b.kot]

Hi.

 

After reconfiguration of the client (I set once again server, port and user credentials) the import was partially successful! I have imported the same dump file but this time authors where imported correctly.  It seems that the root of problem  was in the client (Debian Squeeze 32bit, Plastic version 4.1.10.394).
It wasn't total success because, as I mentioned before, import ignores "master" branch - only feature branches were imported but its step in the right direction.
In the mean time, we decided, despite all the problem with migrating our repository, to switch to Plastic SCM. We created empty repository, copied working copy into workspace and checked in. Loosing project history is painful but we hope that PlasticSCM is worth it.

 

Best regards

[manu]

Hi!

 

the thing regarding the missing branch is very strange, notice that by default the branch explorer is having a time filter, maybe it's hidden by the filter?

 

Can you tell me if when the fast-import ends the "/master" branch is listed in the branch list?

 

Can you verify that in the fast-export file there are commits referencing the /master branch?

 

As you are going to start with Plastic SCM did you try the new Git BiSync? It's another was to import data from Git. Here you have how you can set-up a Git server -> http://git-scm.com/book/ch4-2.html

 

(I'm assuming that you are coming from Git)

[b.kot]

Hi.

 

Can you verify that in the fast-export file there are commits referencing the /master branch?

 

 

I am terribly sorry - you were right - the master branch was not in the dump file. (It was lost by svn-all-fast-export but I should had checked this before posting)

For my defence I can only say, that in the smaller export, which I used to determine the problem with importing authors of the commits, the master branch was present.

 

(I'm assuming that you are coming from Git)

Not directly. Git is just a middle step - our source repository is in SVN. So missing master brunch(which should be created from the trunk) and mismatched branches history was caused by svn-all-fast-export. Importing to PlasticSCM just propagated these errors further, but not caused them.

 

Thank you for your attention and once again sorry for the false alarm with missing branch.

 

Best regards

Bartlomiej

[manu]

Hi Bartlomiej,

 

hehehehe, don't worry, sometimes it's hard to realize where the problem is coming from.

 

If you have more questions don't hesitate to contact us!