Re: [RFC PATCH 1/1] mv: integrate with sparse-index

[Derrick Stolee <derrickstolee@github.com>]

On 3/21/2022 3:14 PM, Junio C Hamano wrote:
> Derrick Stolee <derrickstolee@github.com> writes:
> 
>>>> Another tool that may help you here is 'git ls-files --sparse -t'. It lists
>>>> the files in the index and their "tags" ('H' is "normal" tracked files, 'S'
>>>> is SKIP_WORKTREE, etc. [4]), which can help identify when a file you'd
>>>> expect to be SKIP_WORKTREE is not and vice versa.
>>>
>>> Wonderful.
>>>
>>> Quite honestly, because the code will most likely compile correctly
>>> if you just remove the unconditional "we first expand the in-core
>>> index fully" code, and because the "sparse index" makes the existing
>>> index walking code fail in unexpected and surprising ways, I
>>> consider it unsuitably harder for people who are not yet familiar
>>> with the system.  Without a good test coverage (which is hard to
>>> give unless you are familiar with the code being tested X-<), one
>>> can easily get confused and lost.
>>
>> Certainly, 'git mv' is looking to be harder than expected, but there
>> is a lot of interesting exploration happening in the process.
> 
> Yeah, I know.
> 
> I am suprised that it is harder than expected *to* *you*, though.
> After having seen a few other topics, I thought that you should know
> how deceptively easy to lose "require-full" and how hard to audit
> the code that may expect "a flat list of paths" in the in-core index
> ;-).

I'm particularly surprised in how much 'git mv' doesn't work very
well in the sparse-checkout environment already, which makes things
more difficult than "just" doing the normal sparse index things.

It's good that we are discovering them and working to fix them.

Thanks,
-Stolee