Re: [PATCH] sha1_file.c:create_tmpfile(): Fix race when creating loose object dirs

[Junio C Hamano <gitster@pobox.com>]

Johan Herland <johan@herland.net> writes:

>> Agreed.  We may notice the failure to correct the permissions in the
>> new code, where the old code left existing directories incorrect
>> permissions as they were.
>
> I'm trying to mentally construct a scenario where two writers with
> different configuration contexts ...
> This case is unfortunate, but no different than if steps 3 and 4 are
> swapped, i.e. the case where fetch B fails because the repo is not yet
> group-writable. Also, remember that as part of changing
> core.sharedRepository like this (step 3), we also require the admin to
> manually alter the permission bits of existing object dirs, to make
> sure they are group-writable (call this step 3.5). All of this has to
> happen _while_ fetch A is running.
>
> Trying to code around this (frankly insane) case in the
> create_tmpfile()/adjust_shared_perm() code is IMHO silly.

I agree.  Note that "We may notice" above were not meant as "we
would fail unnecessarily in some cases where we did not, which is a
regression".  Both the old and the new code will have problems when
two different users race with each other while having umask that is
unsuitable for working together.