Re: [PATCH v2 2/3] submodule: port submodule subcommand 'add' from shell to C

From: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Date: Nov 19, 2020, 07:44 UTC
Message-ID: <871rgprdt1.fsf@evledraar.gmail.com>
In-Reply-To: <20201118231331.716110-1-jonathantanmy@google.com>

On Thu, Nov 19 2020, Jonathan Tan wrote:

Show 17 quoted lines

>> Whew.
>> 
>> This was way too big to be reviewed in a single sitting.  I do not
>> know offhand if there is a better way to structure the changes into
>> a more digestible pieces to help prevent reviewers from overlooking
>> potential mistakes, though.
>> 
>> Thanks.
>
> I just took a look at this, and one thing that would have helped is if
> you ported the end of the function first in a commit, and work your way
> backwards (in one or more commits).
>
> After reading through the whole thing, I saw that this is mostly a
> straightforward start-to-finish port (besides factoring out code into
> functions), but it would be much easier for reviewers to conceptualize
> and discuss the different parts if they were already divided.

Having done some minor changes to git-submodule.sh recently, I wondered if we weren't at the point where it would be a nice approach to invert the C/sh helper relationship.

I.e. write git-submodule.c, which would be the small entry point, it would then mostly dispatch to a submodule--helper, which would in turn mostly dispatch to a new submodule--helper-sh (containing most of the current git-submodule.sh code), which in turn would re-dispatch to the C submodule--helper (which as an aside, then sometimes calls itself via process invocation).

It's quite a bit of spaghetti code, but means that there's a straighter path to porting some of the setup code such as the "--check-writeable", is_absolute_path() etc. being changed at the start of the change here to git-submodule.sh.