->>> * Would it be better for `rcs_revert` to not commit, and
->>> `rcs_commit_staged` to then be used? This would work for git, but
->>> maybe other RCSs would be problimatic. It would simplifiy the
->>> interface and allow for future mulitple-revert interfaces.
->>> * I quite don't understand why one caller of `git_parse_changes`
->>> needs it to chdir, and not the other one. It's running
->>> in the same git repo either way, and git doesn't need
->>> `git show` to run in a subdir at all..
->>>> I was aping (preserving) what was already there. I don't understand what you say about `git show` - it must run under $srcdir, surely? And empirically the CGI process wasn't in the right place. By all means simplify that. -- [[peteg]]
-
->>> * Probably needs to untaint the revs passed in.
->>> * Seems backwards for `rcs_preprevert` to import and
->>> use `IkiWiki::Receive`.
->>>> Indeed. This is saying that the checking code in IkiWiki::Receive is in the wrong place. I think it would be better to set up some general hooks and shuffle it into a plugin, for then other plugins that maintain special files in the repo can have a say about validity. -- [[peteg]]