diffoscope

Compare two files

$ diffoscope [file1] [file2]

$ diffoscope --html [report.html] [file1] [file2]

$ diffoscope --max-diff-block-lines [100] [file1] [file2]

$ diffoscope [dir1] [dir2]

$ diffoscope --json [report.json] [file1] [file2]

$ diffoscope [package1.deb] [package2.deb]

$ diffoscope --exclude "[*.pyc]" [dir1] [dir2]

diffoscope [options] path1 path2

diffoscope performs deep comparison of files, recursively unpacking archives and examining contents. Unlike simple diff, it understands file formats and shows meaningful differences inside compressed files, images, binaries, and more.
The tool supports hundreds of file formats: archives (tar, zip, deb, rpm), binaries (ELF, PE), images (PNG, JPEG), documents (PDF), and many others. It uses specialized tools to decode each format before comparison.
diffoscope was created to debug reproducible build issues, helping identify why two builds of the same source produce different binaries. Output can be terminal text, HTML for web viewing, or JSON for programmatic processing.

--html file

Write HTML report to file.

Write text report to file.

Write JSON report to file.

Maximum lines per diff block.

Maximum report size.

Exclude files matching pattern.

Ignore directory metadata differences.

Fuzzy matching threshold (0-400).

Treat absent files as empty.

Remove default size limits.

External diff tool to use.

Full functionality requires many external tools (binutils, unzip, etc.). Large files or deep archives may be slow to process. Some proprietary formats aren't supported. Memory usage can be high for large comparisons.

diffoscope was created by the Reproducible Builds project, specifically by Lunar and others, starting in 2014. Originally named "debbindiff" for Debian package comparison, it was renamed and generalized. The tool is central to the reproducible builds effort, helping developers understand why builds differ and fix the sources of non-determinism.

diff(1), cmp(1), reprotest(1)