Re: Incorrect unified diff when run with "--find-copies-harder"

From: Andrei Rybak <rybak.a.v@gmail.com>
Date: Jun 24, 2018, 13:33 UTC
Message-ID: <6d5d2536-0f69-544d-23c7-351e8618ba94@gmail.com>
In-Reply-To: <CAL78Jw0mQWxO4xkzFBoG0Oe4wjomUGj=dP5ufdUrE5Ku0GGEtg@mail.gmail.com>

On 2018-06-24 12:36, Daniel Penkin wrote:

> Hello,
>

Hi,

Show 25 quoted lines

> I believe I found a bug in how Git represents a diff when invoked with
> "--find-copies-harder" parameter.
> Specifically, the unified diff header of a hunk contains an extra
> piece of text which appears to be a line from the context (i.e.
> unchanged line), something like this:
> 
>     > git diff --find-copies-harder d00ca3f 20fb313
>     diff --git a/test.txt b/copy.txt
>     similarity index 81%
>     copy from test.txt
>     copy to copy.txt
>     index 734156d..43a3f9d 100644
>     --- a/test.txt
>     +++ b/copy.txt
>     @@ -2,6 +2,7 @@ line 1
>      line 2
>      line 3
>      line 4
>     +added line
>      line 5
>      line 6
>      line 7
> 
> Note "line 1" after the standard unified diff header.
>

This text after @@ is usually a function name in a programming language or some other relevant part of hunk context, to help user navigate the diff more easily. What you are getting is the default version of it, as it is just comparing txt files. You can read more about it in the documentation of gitattributes:

https://git-scm.com/docs/gitattributes#_defining_a_custom_hunk_header

Show 10 quoted lines

> I prepared a sample repository with a minimal file I can reproduce
> this problem with:
> https://bitbucket.org/dpenkin/find-copies-harder-bug
> 
> I'm running Git 2.18.0 on a macOS, but I also tried with Git 2.15.0
> and 2.8.6 running on Alpine Linux and was able to reproduce the same
> problem.
> 
> Please advise whether this is expected output or is indeed a bug.
>

This is expected output.

Show 5 quoted lines

> Thank you.
> 
> Kind regards,
> Daniil Penkin
>

-- Best regards, Andrei R.