git-svn --ignore-paths

Question

I\'m struggling for a number of hours now with the --ignore-paths option to git-svn, in an attempt to fetch only certain tags from a large repository.

I want to start

Answer 1

I have a similar problem and a partial Solution for my case .. .

Context :
We have only one SVN repository for Meca, Hardware, Software team... the repository is a complete mess.. so I try to use regex to reduce the area to scan. After 1 day I just gave up.

Finally I used the include-path option to scan only folder with "*Src*" inside. which speed up the scan. also use the option :
-r to reduce the history size you will get in local.
--no-minimize-url otherwise git-svn will scan the whole repository even if you specify the trunk and branch location.

git svn clone 
-r11213:HEAD 
--prefix svn/
--no-minimize-url
--trunk=/trunk/dev/SW/Code/Controller1
--branches=/branches/SW_team/
--include-paths=.*Src.*
https://svnserver.compagny.com/Project1/
Controller1__git__

notice that right now i do not care of the Tags.

Hope it could help, even It's not the original question (5 years ago :-) )

---

EDIT: I cannot add a comment so I comment the question here (not enough reputation point)

1) --ignore-paths can be given for git svn [init/fetch or clone] (i do not know if there is a different behavior)
2) --ignore-paths expect a regex , be carefull the "." means any character. By chance the carater "." is also any character so regex=Acme-5.0 will match string="Acme-5.0" but also string="Acme-580", it should work anyway.

Answer 2

I have been experiencing strange problems with --ignore-paths too. git-svn seems to ignore the entire regexp in some cases. I have seen the same regexp working on repos 1 and ignored on repos 2, where both repos have the same file structure, but different history.

Although I don't see anything wrong with your regexp for your specific tree, I would recommend using the ^ caret at the beginning to specify the ignored paths starting at the root. This might help the rexexp parser speed-up the search and avoid issues where a match could also be found deep inside the trunk for example.

I would use something like --ignore-paths="^tags/Acme-(4|5.[0-4])"

Answer 3

I've struggled with the exact same problem and started editing .git/config to explicitly list the branches or tags that I want.

That approach worked well until I came across a svn repository with lots of branches, so I duly added the ones I wanted and left out the ones I did not. But this failed with configuration file errors. Trial and error apparently shows that there is a limit either the number of branches in the config file or more likely the total number of characters between the opening { and closing }.

My life would be much easier if I could just build regexs.

Answer 4

I was having this same problem today: my regexp would just never match... Make sure you know what the target paths actually look like. I was making an incorrect assumption about the structure of the paths that were being fed to my regexp.

To find out what the paths look like, make git-svn output each path to the console as it tests them:

NOTE: Just in case, make a backup copy of the git-svn file first!

1. Open the git-svn script in a text editor. My script was <git-dir>/libexec/git-core/git-svn.
2. Locate the is_path_ignored subroutine.
3. Add a print statement above the first return statement, as follows...

sub is_path_ignored {
    my ($self, $path) = @_;

    print STDERR "$path\n"; //<-- **ADD THIS LINE**

    return 1 if in_dot_git($path);
    return 1 if defined($self->{ignore_regex}) &&
            $path =~ m!$self->{ignore_regex}!;
    return 0 unless defined($_ignore_regex);
    return 1 if $path =~ m!$_ignore_regex!o;
    return 0;
}

Now use git-svn again with the --ignore-paths switch.

I realised that instead of paths like trunk/baz it was actually using bar/trunk/baz

So instead of

--ignore-paths='^(?:trunk|branches|tags)/baz' 

I needed to use

--ignore-paths='^bar/(?:trunk|branches|tags)/baz'

Don't forget to remove the print statement from the git-svn script.

Answer 5

I'm posting this for everyone, who was aslo trying to use --ignore-paths for fetching only specific branches/tags...

After a while struggling with --ignore-paths, which resulted in the following pattern to ignore all folders in branches folder, except folder branchname1 and branchname2:

--ignore-paths='branches/(?!branchname1|branchname2)'

Howerver, the correct solution is hiding at the bottom of the GIT SVN documentation:

It is also possible to fetch a subset of branches or tags by using a comma-separated list of names within braces. For example:

[svn-remote "huge-project"]
  url = http://server.org/svn
  fetch = trunk/src:refs/remotes/trunk
  branches = branches/{red,green}/src:refs/remotes/project-a/branches/*
  tags = tags/{1.0,2.0}/src:refs/remotes/project-a/tags/*

So in your case, .git/config should contain something like this:

tags = tags/{Acme-4.x,Acme-5.0,Acme-5.1,Acme-5.2,Acme-5.3,Acme-5.4}:refs/remotes/origin/tags/*

Answer 6

You might also just try:

cat .git/config

on Linux, or:

type .git\config

on Windows, from your new repository directory, to see the fetch URL, the branches and tags url.