On 10/8/2019 11:51 PM, wuzhouhui wrote:
>> -----Original Messages-----
>> From: "Junio C Hamano" <gits...@pobox.com>
>> Sent Time: 2019-10-09 11:02:44 (Wednesday)
>> To: wuzhouhui <wuzhouhu...@mails.ucas.ac.cn>
>> Cc: git@vger.kernel.org, cuif...@sugon.com
>> Subject: Re: How to find the commit that erase a change
>> wuzhouhui <wuzhouhu...@mails.ucas.ac.cn> writes:
>>> I have a file which contains complicated change history. When I use
>>>     git log -p file
>>> to see all changes made in this file, I found that a change disappeared
>>> for no reason.
>> "git log [-p] <pathspec>" is not about seeing *all* changes made to
>> the path(s) that match the pathspec.  Especially when your history
>> has merges, the command is to give you _one_ simplest explanation as
>> to how the contents of the path(s) came to be in the shape you see
>> in HEAD.
>> So for example, if you have a history like this (time flows from
>> left to right):
>>     O-----A-----B----M-----N
>>            \        /
>>             \      /
>>              X----Y
>> where A or B did *not* touch "file", X added a definition of func()
>> to "file", Y reverted the change X made to "file", M made a natural
>> merge between B and Y and N did not touch "file", "git log N file"
>> would not even show the existence of commits X or Y.  In the larger
>> picture, at ancient time O, the file started without func(), and
>> none of the commits A, B, M or N felt the need to add it and as the
>> result, N does not need the unwanted func().  So "file's contents
>> are the same since O throughout the history reaching N" is given as
>> _one_ simplest explanation.
>> The "--full-history" option may help, though.
> "--full-history" doesn't resolve my problem, but
>     git log -p -c file
> does. I found that my change was erased in a merge commit.

In these cases of erased merge commits, I find that

        git log --simplify-merges -- file

works best for finding the merge responsible. Just the
--full-history option may include many extra merges (if you
work in a repo with many collaborators).


Reply via email to