cloc --git --diff fails to identify "same" files and code from v2.04 on

[dennisbrendel]

Describe the bug
cloc v2.04 broke the --git--diff accounting of the "same" lines of code. This was noticed in v2.06 but it is already present in v2.04.

cloc; OS; OS version

• cloc version: v2.04
• If running the cloc source, Perl version: v5.42.0
• OS (eg Linux, Windows, macOS, etc): Fedora
• OS version: 43

To Reproduce

mkdir cloc_reproducer
cd cloc_reproducer
git init
echo "int main(void) {}" > file.c
git add .
git commit -am "initial commit"
git tag 1.0
echo "file.c" > list.txt
git add list.txt
git commit -m "Add list file"
git tag 2.0

Now run cloc

cloc --diff --git --list-file list.txt 1.0 2.0
       1 text file.
       1 text file.
       0 files ignored.                             
Nothing to count.

Expected result

cloc --diff --git --list-file list.txt 1.0 2.0
       1 text file.
       1 text file.
       0 files ignored.                             

github.com/AlDanial/cloc v 2.02  T=0.06 s (16.6 files/s, 16.6 lines/s)
-------------------------------------------------------------------------------
Language                     files          blank        comment           code
-------------------------------------------------------------------------------
C
 same                            1              0              0              1
 modified                        0              0              0              0
 added                           0              0              0              0
 removed                         0              0              0              0
-------------------------------------------------------------------------------
SUM:
 same                            1              0              0              1
 modified                        0              0              0              0
 added                           0              0              0              0
 removed                         0              0              0              0
-------------------------------------------------------------------------------

Additional context
Add any other context about the problem here.

[AlDanial]

Thanks for the reproduction steps; they duplicated the problem on my computer.

The git diff logic changed a lot (for the better, mostly) in 2.04, primarily to be able to compare files whose names changed between the comparison versions. One of the new steps involves calling git --diff-filter=ADRM. Using your inputs the results would be

» git -c "safe.directory=*" --no-pager diff --name-status --diff-filter=ADRM 1.0 2.0
A       list.txt

From this output, if cloc were running properly, it would say one file (list.txt) was added. It won't report anything for "same" since the git output is silent on unchanged files (file.c in this case). I'll figure out why there's no output, but the output that appears won't match your expected result.

[AlDanial]

The issue is the only entry in list.txt is the file git will ignore; it prevents cloc from having anything to count. If you add list.txt to itself you'll get

-------------------------------------------------------------------------------
Language                     files          blank        comment           code
-------------------------------------------------------------------------------
Text
 same                            0              0              0              0
 modified                        0              0              0              0
 added                           1              0              0              1
 removed                         0              0              0              0
-------------------------------------------------------------------------------
SUM:
 same                            0              0              0              0
 modified                        0              0              0              0
 added                           1              0              0              1
 removed                         0              0              0              0
-------------------------------------------------------------------------------

[dennisbrendel]

Thanks @AlDanial .

I'm using this functionality in a larger project, namely the Linux kernel, with long lists of source files, where it is important to count the changed files as well as the unchanged files. So I need to account for the "same" files.

From the git documentation it seems like using the --diff-filter option is not able to account for unchanged files.

[AlDanial]

cloc v 2.02 does git diff file pairing the old way--directly by filenames--instead of using git itself to identify the pair and therefore gives the answer you're looking for:

cloc-2.02.pl --diff --git --list-file list.txt 1.0 2.0
       1 text file.
       1 text file.
       0 files ignored.                             

github.com/AlDanial/cloc v 2.02  T=0.07 s (15.3 files/s, 15.3 lines/s)
-------------------------------------------------------------------------------
Language                     files          blank        comment           code
-------------------------------------------------------------------------------
C
 same                            1              0              0              1
 modified                        0              0              0              0
 added                           0              0              0              0
 removed                         0              0              0              0
-------------------------------------------------------------------------------
SUM:
 same                            1              0              0              1
 modified                        0              0              0              0
 added                           0              0              0              0
 removed                         0              0              0              0
-------------------------------------------------------------------------------

I'm guessing you already know this based on your initial comments. Now I'm scratching my head trying to think of a way to have both unchanged files and track renames.

[bulwahn]

Along a similar setup, the response "Nothing to count." is bad to parse by further tools, especially when passing the --json option, which means the parser expects a json file but gets something different.

For the setup described above by Dennis, try to run 'cloc --quiet --sum-one --json --diff --git --list-file list.txt 1.0 2.0'.

This results in the response "Nothing to count." in cloc-2.06, which is not a valid json object, whereas cloc-2.02 would return a valid json object.