Re: Re: [PATCH v3] builtin/clone.c: add --reject-shallow option

From: lilinchao@oschina.cn <lilinchao@oschina.cn>
Date: Feb 20, 2021, 10:40 UTC
Message-ID: <1689e980736811eb80250026b95c99cc@oschina.cn>
In-Reply-To: <eaa219a86bbc11ebb6c7a4badb2c2b1165032@pobox.com>

-------------- lilinchao@oschina.cn

Show 37 quoted lines

>"lilinchao@oschina.cn" <lilinchao@oschina.cn> writes:
>
>> Sorry, you may forget there is a re-read git-config under these lines
>
>Ahhhh, thanks.  You're right that "clone" is "special" in that way
>and requires unusual order---I forgot about that.
>
>>>This is only for cloning from a local repository, no?  IOW, path at
>>>this point may even be "git://github.com/git/git/" and checking with
>>>access() does not make sense.
>>>
>>>Ah, it is even worse.  get_repo_path() can return NULL, so mkpath()
>>>will crash in such a case.  This must be at least
>>>
>>>	if (path && !access(mkpath("%s/shallow", path), F_OK))
>>>	is_shallow = 1;
>>> ...
>>>
>>>So, I think the above two hunks are making the code worse.  If we
>>>are to detect and reject cloning from the shallow repository when
>>>going through the transport layer (i.e. "--no-local" or cloning from
>>>"git://github.com/git/git", or "https://github.com/git/git", if it
>>>were a shallow repository), that must be handled separately.
>>>
>> Sorry, I made the question simple. Reject cloning a shallow repository
>> should apply to all four type transport protocols. There still a bunch of
>> work to be done.
>
>Sorry, but I didn't realize that this is just a work-in-progress
>that shows an early "only local transport is covered" preview.  I
>think, modulo that the access(mkpath()) thing should be in the "if
>(is_local)" block, the patch covers the case of local transport well
>enough.  I think by "four types of transport" you mean ssh://, git://
>https:// etc., but they all go through the same transport API, so
>hopefully once one of them, say, git://, is covered, that would take
>us a long way.
>

Hi, after some exploration, I found that the following changes
can achieve the goal:
(around at line 1386 in builtin/clone.c)
    /* below part is my changes */
            if (reject_shallow) {
                    if (local_shallow || is_repository_shallow(the_repository)) {
                            die(_("source repository is shallow, reject to clone."));
                    }
            }
    /* end */
            update_remote_refs(refs, mapped_refs, remote_head_points_at,
                               branch_top.buf, reflog_msg.buf, transport,
                               !is_local);

as it shows, using existing method "is_repository_shallow" to detect whether the repo is shallow or not, before "update_remote_refs" I made some tests based on this change, and it seems ok. what do you think about this?

>Thanks.