repomapper(1) update and manipulate contributor maps


repomapper [-i] [-p passwordfile] [-u updatefile] [-h host] contribmap


Older, centralized version-control systems such as CVS and SVN centralize a repository on one host and identify users by their account names on that host. Distributed version-control systems such as git and Mercurial identify users by a netwide-unique ID consisting of a name-among-humans followed by an email address.

When moving a repository from a centralized to a distributed system, therefore, one of the prerequisites is a contributor map that associates each account name on the old system to a DVCS-style ID on the new one. This tool automates parts of that process.

The main argument file must be a contributor map such as is read by the authors read subcommand of reposurgeon(1). It may be a fresh or stub map, produced by authors write before any human-name or email information has been added to the repository. Or it may have name-among-humans and email information filled in for some entries.

A stub map entry looks something like this:

foonly = foonly <foonly>

The same entry, fully filled in, might look something like this:

foonly = Fred Foonly <[email protected]>

The default behavior of the tool is to report all map entries, in effect a sorting copy of the file.

With -i, it reports only entries that are not yet in DVCS form - that is, either the fullname field on the right side of the equals sign is identical to the account name on the left, or the email field contains no @-sign, or both.

With the -p option, this tool fills in the full-name field using the password file given as the option's argument. Only the username and the comment (or 'gecos') field containing the user's name-among-humans are used. Other fields are ignored, including the password-hash field. (On modern Unixes this field does not contain the actual hash, which lives in a different file named /etc/shadow, so /etc/passwd can be shared without security risk.)

In the -p mode, for each entry in the contrib file the program looks for a username in the password file matching the name to the left of the equal sign. If a match is found, the user's name-among-humans is extracted from the gecos field and replaces the text between the "=" and the "<".

Thus, the stub line above and the /etc/passwd line

foonly:x:1000:1000:Fred Foonly,,,:/home/foonly:/bin/bash

will combine to produce this on output:

foonly = Fred Foonly <foonly>

Note that the email-address part (and, if present, the optional trailing timezone field) are not normally modified.

However, if the -h option is given, the argument is taken to be a host name which should be appended (after a @) to every email field that does not already contain a @. The argument would typically be the fully-qualified domain name of the repository host.

Thus, if the passwd file still contains an entry for every committer (which might not be the case if inactive committer accounts were ever removed), -p mode combined with an -h option can produce an entire, valid contributor map.

In the -u mode of operation, the option argument must be a second contributor file, which is taken as a source of updates. Each contributor entry with a username not matching any in the first contributor map is copied into the first map, which is output.

Output from this tool is always a contrib map sorted by username.


Eric S. Raymond <[email protected]>. This tool is distributed with reposurgeon; see the project page at m[blue][].