On Tue, Jul 18, 2017 at 3:32 PM, Junio C Hamano <gitster@xxxxxxxxx> wrote:
> Stefan Beller <sbeller@xxxxxxxxxx> writes:
>
>>> + if (!lstat(list_item->name, &st) && !ce_match_stat(list_item, &st, 0)) {
>>> + print_status(info, ' ', list_item->name, sub_sha1, displaypath);
>>
>> The question from the last round still stands
>> https://public-inbox.org/git/CAGZ79kb18z5zc9iu3Vv5aVZWJmoZzmwbMVpy89VC-t-ei2M+bw@xxxxxxxxxxxxxx/
>>
>> I am not an expert in the diff area and wonder how
>> the cmd_diff_files functionality is achieved with just a stat call
>> and then comparing it to ce_match_stat. 'Using "dirty" ignores
>> all changes to the work tree of submodules, only changes to the
>> commits stored in the superproject are shown.' So I'd have
>> expected ce->oid to be compared (is there an index entry differing,
>> i.e. more than one stage?)
>
> ce_match_stat() calls into ce_compare_gitlink() for a 160000 entry,
> which would resolve HEAD ref there and compares ce->oid with it.
Oh in that case this should be fine, as in the original we did
"git diff-files --ignore-submodules=dirty <path>",
which did precisely that.
> But as you said, this is probably insufficient to emulate the
> original. Shouldn't it call into run_diff_files(), which is the
> in-core way to run the equivalent of "diff-files"?
Oh, your comment in [1] was related to cmd_diff_files,
which is more complicated than run_diff_files?
run_diff_files also iterates over all cache entries.
I think we need to be looking at match_stat_with_submodule
to figure out what we need to do.
[1] https://public-inbox.org/git/xmqq60fdoyyt.fsf@xxxxxxxxxxxxxxxxxxxxxxxxxxx/
