Re: git rev-list fails to verify ssh-signed commits (but git log works)

Jeff King [off-list ref] writes:

I'd be more worried about correctness (command "foo" must not parse
option "bar" because it is plumbing and must use the default). But
looking over the options, I really don't see anything like that. The one
I'd expect (or worry most about) is "we do/do not bother to enable
signatures at all based on config" but I don't think that is the case.
We default use_format to &gpg_format[0], so there is always a signature
format defined, even if the config is skipped.

I arrived at the same conclusion after seeing what these
configuration affected.  If some plumbing codepaths should avoid
certain gpg-interface features, I didn't see a good reason for them
to be calling into gpg-interface.c functions (e.g. that would have
meant that "rev-list '--format=%G'" should have errored out with a
"%G is only available in log but not in rev-list which is a
plumbing" error message) and at that point reading configuration
is much lessor problem.

If the gpg code used git_config_get_string(), etc, then they could just
access each key on demand (efficiently, from an internal hash table),
which reduces the risk of "oops, we forgot to initialize the config
here". It does probably mean restructuring the code a little, though
(since you'd often have an accessor function to get "foo.bar" rather
than assuming "foo.bar" was parsed into an enum already, etc). That may
not be worth the effort (and risk of regression) to convert.

Yup.  Care must be taken not to break "-u <keyid>" that is set via
calling set_signing_key() directly, which assumes that the default
key is already set from the config earlier, but the resulting code
may make it clearer that our rule is "command line first, and then
peek into configuration as a fallback value", if we did such a
conversion.

Thanks.