Compare Dosbox DRO v2 files

Quick and easy CLI tool to compare 2 Dosbox DRO v2 files.

dro_compare example_files/0.dro example_files/1.dro

This tool compares the headers, as well as any commands send to the registers. It has optional handling of delays (see below). If no arguments are specified it goes through the file, matching every command and value send to the registers.

Optionally it can discard any delays in the songs. This is usefull if you want to compare 2 different recordings, since small millisecond differences can occur in delays between the recording.

dro_compare -r example_files/0.dro example_files/1.dro

This will add up any delays found in the recording, and validate if the total match. For example 2x1ms delay would still match a 1x2ms delay. Note that this does not guarantee 100% correctness of timing within the recording, just that the same amount of delays are there across the song.

dro_compare -m example_files/0.dro example_files/1.dro

Only Linux, GCC and some standard C libraries are required. Simply run:

make install

and it should be installed in /usr/local/bin

Small millisecond delays show up in the file as command/value pairs, but not always at the same location or exact duration, so comparing them is hard. The exact cause of these delays differences is hard to pinpoint, but is assumed to be mainly because Dosbox and derivatives are not cycle-accurate (source).

Dosbox is not 100% accurate in it's emulation, so small discrepancies can sneak into an audio recording, throwing off most comparison techniques. For example taking recording 0.wav and 1.wav from the example_files folder and comparing them shows minute differences. You won't hear them even with the best audio equipment, but they are there, sometimes differences as small as a few samples/milliseconds. This is not just from the start of the actual recording, which can vary in the bit of silence in between, but also during the song these small differences are observed.

I've tried many different approaches to try and match the audio, with NO success:

Importing 2 audio waveforms, truncating the silence at the start, inverting the second waveform and then listening to the result. In a 100% duplicate recording the net result should be complete silence, since these 2 waveforms cancel each other out.

This tool can compare audio , correct for level differences, phase correction, clock drift correction etc. Also this tool reports the recording are not the same.