Re: [PATCH v2 11/14] config: pass kvi to die_bad_number()

From: Glen Choo <chooglen@google.com>
Date: Jun 2, 2023, 17:23 UTC
Message-ID: <kl6lpm6dzskk.fsf@chooglen-macbookpro.roam.corp.google.com>
In-Reply-To: <20230601234829.432150-1-jonathantanmy@google.com>

Jonathan Tan <jonathantanmy@google.com> writes:

Show 13 quoted lines

> "Glen Choo via GitGitGadget" <gitgitgadget@gmail.com> writes:
>> Outside of config.c, config callbacks now need to pass "kvi" to any of
>> the git_config_<type>() functions that parse a config string into a
>> number type. Included is a .cocci patch to make that refactor. In cases
>> where "kvi" would never be used, pass NULL, e.g.:
>> 
>> - In config.c, when we are parsing a boolean instead of a number
>> - In builtin/config.c, when calling normalize_value() before setting
>>   config to something the user gave us.
>
> In these cases, could we synthesize a kvi instead of using NULL? I
> believe there are already code paths that use an UNKNOWN scope - these
> seem similar to that.

Okay, that sounds reasonable. This has echoes of Philip Wood's suggestion (elsewhere in the thread) of combining all of the config_fn_t args into a single struct, which means we can no longer use NULL as the default.