Phillip Wood <phillip.wood123@xxxxxxxxx> writes: >>> @@ -957,7 +976,11 @@ static int run_git_commit(struct repository *r, >>> if (opts->committer_date_is_author_date) >>> argv_array_pushf(&cmd.env_array, "GIT_COMMITTER_DATE=%s", >>> + opts->ignore_date ? >>> + "" : >>> author_date_from_env_array(&cmd.env_array)); >>> + if (opts->ignore_date) >>> + argv_array_push(&cmd.env_array, "GIT_AUTHOR_DATE="); >>> >> >> Perhaps this could be done in read_env_script(), too, instead of fixing >> up what this function did right after calling it, twice? Something like >> this: > > Long term I'd like to stop writing the author-script file unless we're > stopping for a conflict resolution or edit so I'm reluctant to add > more functionality to it at the moment. Also I tend to view > read_env_script() it as loading the author details from a file rather > than setting up the environment for the commit. Yeah, the author-script, once the program halts and returns the control to the human-user sitting on the terminal, can be peeked at, and it should record what the program learned from the commit; it smells like a premature optimization to rewrite it to what the program plans to use to replay the commit and lose the information taken from the original.
