cloc --git-diff-rel does not respond consistently when dealing with copy/rename

[EnricoPicci]

Describe the bug
The command cloc --git-diff-rel --csv --by-file commit-sha^1 commit-sha does not always return the same number of records, i.e. the same number of differences.
In particular, differences which are marked as copy/rename by the git diff --numstat, sometimes are skipped by cloc --git-diff-rel and sometimes are listed as 2 differences.

cloc; OS; OS version

• cloc version: 1.98
• If running the cloc source, Perl version:
• OS Linux:
• OS version: 22.04

To Reproduce
I have tried to reproduce the issue on some public repos but I have not been able to reproduce it

Expected result
Every run of cloc --git-diff-rel --csv --by-file commit-sha^1 commit-sha should return the same amount of records.

Additional context
Let's say that git diff --numstat returns something like
1 1 src/{old-subdir=> new-subdir}/my-file.java
stating that my-file.java has been moved from old-subdir to new-subdir adding and removing one line of code.

what I get if I run the same comparison with cloc --git-diff-rel --csv --by-file what I get is any of these responses:

• src/new-subdir/my-file.java has one line removes and one line changed, no mention to the file src/old-subdir/my-file.java
• src/new-subdir/my-file.java has 100 lines added and src/old-subdir/my-file.java has 100 lines removes (assuming that the files contains 100 lines)

which one of the 2 answers I get is random.

[AlDanial]

This problem will be difficult (impossible?) to resolve without a way for me to reproduce it. For now I will close this issue but please reopen it if you find a public repo that can demonstrate the inconsistent output.

In the long term, I have a goal to implement in cloc the swith --git-diff-simindex which will attempt to follow renames via git's --find-renames option.

[EnricoPicci]

I think I have found a way to reproduce the problem I have described in the issue.
Clone the repo https://github.com/EnricoPicci/git-metrics.
Then run repeatedly the following command:
cloc --git-diff-rel --csv --by-file --timeout=10 --quiet 6fb8624bad8d62ee14da5c7a527c786b301f7529^1 6fb8624bad8d62ee14da5c7a527c786b301f7529

What I get on my machine (running Ubuntu 22.04.2) is either a list of 11 differences or (less often) a list of 16 differences.

I have written this nodejs script to run the commands few times - the scripts exits when it finds that the result of the previous run of the command is different from the result of the last run of the command

var child_process = require("child_process");
var fs = require("fs");

var previousLength = 0;
var i = 0;
var maxNumberOfIterations = 100;

setInterval(function () {
    console.log('iteration ' + i);
    console.log('lastLength ' + previousLength);
    child_process.execSync('cloc --git-diff-rel --csv --by-file --timeout=10 --quiet  6fb8624bad8d62ee14da5c7a527c786b301f7529^1 6fb8624bad8d62ee14da5c7a527c786b301f7529 > out.csv');
    var lastLength = fs.readFileSync('out.csv', 'utf8').split('\n').length;
    if (previousLength && lastLength !== previousLength) {
        console.log('the command has not returned the same number of lines: once it was ' + previousLength + ' and now it is ' + lastLength);
        process.exit(1);
    }
    if (i === maxNumberOfIterations) {
        console.log('the command has returned the same number of lines every time after ' + i + ' iterations');
        process.exit(0);
    }
    i++;
    previousLength = lastLength;
}, 100);

Attached below you may find both the results I get, the one that returns 11 records and the one that returns 16 records.

out_11.csv
out_16.csv