diffoscope man page

diffoscope — in-depth comparison of files, archives, and directories


diffoscope --help
diffoscope [OPTIONS] [--json output_diff] path1 path2
diffoscope [OPTIONS] diff
diffoscope [OPTIONS] < diff


diffoscope will try to get to the bottom of what makes files or directories different. It will recursively unpack archives of many kinds and transform various binary formats into more human readable form to compare them. It can compare two tarballs, ISO images, or PDF just as easily.

It can be scripted through error codes, and a report can be produced with the detected differences. The report can be text or HTML. When no type of report has been selected, diffoscope defaults to write a text report on the standard output.

diffoscope was initially started by the "reproducible builds" Debian project and now being developed as part of the (wider) .Reproducible Builds. initiative.  It is meant to be able to quickly understand why two builds of the same package produce different outputs. diffoscope was previously named debbindiff.

See the Command-Line Examples section further below to get you started, as well as more detailed explanations of all the command-line options. The same information is also available in /usr/share/doc/diffoscope/README.rst or similar.

positional arguments


First file or directory to compare. If omitted, tries to read a diffoscope diff from stdin.


Second file or directory to compare. If omitted, no comparison is done but instead we read a diffoscope diff from path1 and will output this in the formats specified by the rest of the command line.

optional arguments


Display debug messages


Open the Python debugger in case of crashes

--status-fd FD

Send machine-readable status to file descriptor FD

--progress, --no-progress

Show an approximate progress bar. Default: yes if stdin is a tty, otherwise no.


Disable most default output limits and diff calculation limits.

output types


Write plain text output to given file (use - for stdout)

--text-color WHEN

When to output color diff. WHEN is one of {never, auto, always}. Default: auto, meaning yes if the output is a terminal, otherwise no.


If there was no difference, then output an empty diff for each output type that was specified. In --text output, an empty file is written.


Write HTML report to given file (use - for stdout)

--html-dir OUTPUT_DIR

Write multi-file HTML report to given directory

--css URL

Link to an extra CSS for the HTML report

--jquery URL

URL link to jQuery, for --html and --html-dir output. If this is a non-existent relative URL, diffoscope will create a symlink to a system installation. (Paths searched: /usr/share/javascript/jquery/jquery.js.) If not given, --html output will not use JS but --htmldir will if it can be found; give "disable" to disable JS on all outputs.


Write JSON text output to given file (use - for stdout)

--markdown OUTPUT_FILE

Write Markdown text output to given file (use - for stdout)

--restructured-text OUTPUT_FILE

Write RsT text output to given file (use - for stdout)

--profile OUTPUT_FILE

Write profiling info to given file (use - for stdout)

output limits

--max-text-report-size BYTES

Maximum bytes written in --text report. (0 to disable, default: 0)

--max-report-size BYTES

Maximum bytes of a report in a given format, across all of its pages. Note that some formats, such as --html, may be restricted by even smaller limits such as --max-page-size. (0 to disable, default: 41943040)

--max-diff-block-lines LINES

Maximum number of lines output per unified-diff block, across all pages. (0 to disable, default: 1024)

--max-page-size BYTES

Maximum bytes of the top-level (--html-dir) or sole (--html) page. (default: 409600, remains in effect even with --no-default-limits)

--max-page-size-child BYTES

In --html-dir output, this is the maximum bytes of each child page (default: 204800, remains in effect even with --no-default-limits)

--max-page-diff-block-lines LINES

Maximum number of lines output per unified-diff block on the top-level (--html-dir) or sole (--html) page, before spilling it into child pages (--html-dir) or skipping the rest of the diff block. Child pages are limited instead by --max-page-size-child. (default: 128, remains in effect even with --no-default-limits)

diff calculation


Treat absent files as empty

--exclude GLOB_PATTERN

Exclude files that match GLOB_PATTERN. Use this option to ignore files based on their names.

--exclude-command REGEX_PATTERN

Exclude commands that match REGEX_PATTERN. For example '^readelf.*\s--debug-dump=info' can take a long time and differences here are likely secondary differences caused by something represented elsewhere. Use this option to disable commands that use a lot of resources.

--exclude-directory-metadata, --no-exclude-directory-metadata

Exclude directory metadata. Useful if comparing files whose filesystem-level metadata is not intended to be distributed to other systems. This is true for most distributions package builders, but not true for the output of commands such as `make install`. Metadata of archive members remain un-excluded. Use this option to ignore permissions, timestamps, xattrs etc. Default: False if comparing two directories, else True. Note that "file" metadata actually a property of its containing directory, and is not relevant when distributing the file across systems.

--fuzzy-threshold FUZZY_THRESHOLD

Threshold for fuzzy-matching (0 to disable, 60 is default, 400 is high fuzziness)

--tool-prefix-binutils PREFIX

Prefix for binutils program names, e.g. "aarch64-linux-gnu-" for a foreign-arch binary or "g" if you're on a non-GNU system.

--max-diff-input-lines LINES

Maximum number of lines fed to diff(1) (0 to disable, default: 4194304)

--max-container-depth DEPTH

Maximum depth to recurse into containers. (Cannot be disabled for security reasons, default: 50)

--max-diff-block-lines-saved LINES

Maximum number of lines saved per diff block. Most users should not need this, unless you run out of memory. This truncates diff(1) output before emitting it in a report, and affects all types of output, including --text and --json. (0 to disable, default: 0)


Automatically use corresponding -dbgsym packageswhen comparing .deb files. (default: False)


Force recursing into the depths of file formats even if files have the same content, only really useful for debugging diffoscope. Default: False

information commands

--help, -h

Show this help and exit


Show program's version number and exit

--list-tools [DISTRO]

Show external tools required and exit. DISTRO can be one of {arch, debian, FreeBSD}. If specified, the output will list packages in that distribution that satisfy these dependencies.


List packages needed for Debian in 'substvar' format.

--list-missing-tools [DISTRO]

Show missing external tools and exit. DISTRO can be one of {arch, debian, FreeBSD}. If specified, the output will list packages in that distribution that satisfy these dependencies.

File renaming detection based on fuzzy-matching is currently disabled. It can be enabled by installing the "tlsh" module available at https://github.com/trendmicro/tlsh

file formats supported

Android APK files, Android boot images, Berkeley DB database files, ColorSync colour profiles (.icc), Coreboot CBFS filesystem images, Dalvik .dex files, Debian .buildinfo files, Debian .changes files, Debian source packages (.dsc), Device Tree Compiler blob files, ELF binaries, FreeDesktop Fontconfig cache files, FreePascal files (.ppu), GHC Haskell .hi files, GIF image files, GNU R Rscript files (.rds), GNU R database files (.rdb), Gettext message catalogues, Git repositories, Gnumeric spreadsheets, Gzipped files, ISO 9660 CD images, JPEG images, JSON files, Java .class files, JavaScript files, LLVM IR bitcode files, LZ4 compressed files, MacOS binaries, Microsoft Windows icon files, Microsoft Word .docx files, Mono 'Portable Executable' files, OCaml interface files, Ogg Vorbis audio files, OpenOffice .odt files, OpenSSH public keys, OpenWRT package archives (.ipk), PDF documents, PGP signed/encrypted messages, PNG images, PostScript documents, RPM archives, Rust object files (.deflate), SQLite databases, SquashFS filesystems, TrueType font files, WebAssembly binary module, XML binary schemas (.xsb), XML files, XZ compressed files, ar(1) archives, bzip2 archives, character/block devices, cpio archives, directories, ext2/ext3/ext4/btrfs/fat filesystems, statically-linked binaries, symlinks, tape archives (.tar), tcpdump capture files (.pcap) and text files.

Exit Status

Exit status is 0 if inputs are the same, 1 if different, 2 if trouble.

Command-Line Examples

To compare two files in-depth and produce an HTML report, run something like:

$ diffoscope --html output.html build1.changes build2.changes

diffoscope will exit with 0 if there's no differences and 1 if there are.

diffoscope can also compare non-existent files:

$ diffoscope /nonexistent archive.zip

To get all possible options, run:

$ diffoscope --help

If you have enough RAM, you can improve performance by running:

$ TMPDIR=/run/shm diffoscope very-big-input-0/ very-big-input-1/

By default this allowed to use up half of RAM; for more add something like:

tmpfs   /run/shm    tmpfs   size=80%    0   0

to your /etc/fstab; see man mount for details.

External Dependencies

diffoscope requires Python 3 and the following modules available on PyPI: libarchive-c, python-magic.

The various comparators rely on external commands being available. To get a list of them, please run:

$ diffoscope --list-tools


Lunar, Reiner Herrmann, Chris Lamb, Mattia Rizzolo, Ximin Luo, Helmut Grohne, Holger Levsen, Daniel Kahn Gillmor, Paul Gevers, Peter De Wachter, Yasushi SHOJI, Clemens Lang, Ed Maste, Joachim Breitner, Mike McQuaid. Baptiste Daroussin, Levente Polyak.


Please report bugs and send patches through the Debian bug tracking system against the diffoscope package: <https://bugs.debian.org/src:diffoscope>

For more instructions, see CONTRIBUTING.rst in this directory.

Join the users and developers mailing-list: <https://lists.reproducible-builds.org/listinfo/diffoscope>

diffoscope website is at <https://diffoscope.org/>


diffoscope is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

diffoscope is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with diffoscope.  If not, see <https://www.gnu.org/licenses/>.

See Also

Referenced By

debdiff(1), debrepro(1).

August 2019 diffoscope 111