On Tue, Sep 28, 2021 at 02:56:03PM +0200, Ævar Arnfjörð Bjarmason wrote: > The git_config_key_is_valid() function got left behind in a > refactoring in a9bcf6586d1 (alias: use the early config machinery to > expand aliases, 2017-06-14), > > It previously had two users when it was added in 9e9de18f1ad (config: > silence warnings for command names with invalid keys, 2015-08-24), and > after 6a1e1bc0a15 (pager: use callbacks instead of configset, > 2016-09-12) only one remained. > > By removing it we can get rid of the "quiet" branches in this > function, as well as cases where "store_key" is NULL, for which there > are no other users. This was the part that I paid most attention to. The only caller which passed "quiet == 1" was git_config_key_is_valid(), which is gone. So all of those "if (!quiet)" conditionals can go away. And the remaining callers all pass a non-null pointer to store_key, so those guard can go away too. > Out of the 5 callers of git_config_parse_key() only one needs to pass > a non-NULL "size_t *baselen_", so we could remove the third parameter > from the public interface. I did not find that potential > simplification to be worthwhile. But the guard for baselen being NULL needs to stay, and it does. I agree with you (and Peff) that stopping here is fine. Thanks, Taylor
