A Large Angry SCM [off-list ref] writes:
Also, git-var complains when used by an account with an empty gcos
field; thereby, breaking all the non-C git commands even when the user
is not committing.
If the _intent_ was to force commiters and author names in commits,
why was the test not placed only in commit-tree.c?
git-var was more or less intentional. Scripts such as
git-applypatch, git-commit and git-tag use the command to grab
COMMITTER_IDENT to generate sign-off line and tagger information
when asked, and commit-tree.c changes alone would not catch
them.
A user eventually would make commit so it may not be a too bad
to _strongly_ encourage setting up these environment variables,
by being nasty ;-). I agree it would be _very_ annoying until
you either fix your gecos and/or environment.
Ideas welcome.