Re: [BUG] Git add <device file> silently fails

From: Alex Riesen <raa.lkml@gmail.com>
Date: Apr 17, 2010, 18:23 UTC
Message-ID: <v2o81b0412b1004171123z20c2c042qabb2b76143390a36@mail.gmail.com>
In-Reply-To: <n2i81b0412b1004171032v713f156ase295cbe7bbedf1f6@mail.gmail.com>

On Sat, Apr 17, 2010 at 19:32, Alex Riesen <raa.lkml@gmail.com> wrote:

Show 17 quoted lines

> On Sat, Apr 17, 2010 at 18:38, Junio C Hamano <gitster@pobox.com> wrote:
>> Alex Riesen <raa.lkml@gmail.com> writes:
>>
>>> I think something like this should make the accident more
>>> noticable:
>>
>> The early skippage done in dir.c (read-directory-recursive) should treat
>> these as ignored just like paths that are ignored with .gitignore
>> mechanism, and if we do so, we shouldn't need this patch to add another
>> codepath to give notification to the user (we would however still need
>> to reword "'add -f' if you really want to add it", though).
>>
>
> I see. Special files are not treated as ignored yet (and there will be
> no way to un-ignore them). I have to read the code for a while,
> ignored pathnames are sometimes stored for later use too.
>

I am tempted to add a mode_t or DT_something to struct dir_entry. Users of read_directory are likely to do an lstat anyway (well, builtin/add.c does), and sometimes there is something to fill it with (get_dtype for DT_UNKNOWN). The only problem is that the structure is allocated a lot...