Compare bzip2 Archives with bzdiff and bzcmp

bzdiff and bzcmp let you diff two bzip2 archives, or an archive against a plain file, without unpacking either one first. You get a clear result: identical, different, or could not be checked. The examples use bzdiff and bzcmp from bzip2 1.0.8, installed here as Ubuntu package version 1.0.8-5.1ubuntu0.1.

Allow about ten minutes. You need a shell, two readable files, and the bzip2 package. These commands read inputs and create only the compressed examples used for testing: no sudo, no change to your source files, no change to services or configuration.

1. Check the installed commands

Confirm the wrappers are available and record the compressor version:

$ command -v bzdiff
/home/linuxbrew/.linuxbrew/bin/bzdiff
$ command -v bzcmp
/home/linuxbrew/.linuxbrew/bin/bzcmp
$ bzip2 --version
bzip2, a block-sorting file compressor.  Version 1.0.8, 13-Jul-2019.

The pathname can differ on your machine. What matters is the program version, because this is a small shell wrapper around diff or cmp, and distributions carry their own packaging changes. Read the installed bzdiff(1) page when a local version differs.

2. Compare two ordinary files

With two arguments, bzdiff accepts compressed or uncompressed files in any combination. Start with two uncompressed files:

$ bzdiff /path/to/old.txt /path/to/new.txt

Silence means the files match, and the command returns status 0. For a readable line-by-line result, pass an option understood by diff:

$ bzdiff -u /path/to/old.txt /path/to/new.txt
--- /path/to/old.txt	2026-09-22 16:00:00.000000000 +0100
+++ /path/to/new.txt	2026-09-22 16:01:00.000000000 +0100
@@ -1,2 +1,2 @@
 alpha
-beta
+gamma

The timestamps and context can differ. The leading minus and plus lines show what changed. A difference is a normal comparison result, not a command failure: diff returns 1 when the files differ and 0 when they match.

3. Compare compressed files directly

The wrapper decompresses inputs as needed and feeds the resulting text to diff, so you can compare two .bz2 files without making permanent uncompressed copies:

$ bzdiff -u /archive/old.txt.bz2 /archive/new.txt.bz2
--- -	2026-09-22 16:27:16.633465530 +0100
+++ /tmp/bzdiff.NrCp3IOfRY	2026-09-22 16:27:16.591432331 +0100
@@ -1,2 +1,2 @@
 alpha
-beta
+gamma
$ printf '%s\n' "$?"
1

Those temporary names in the output are expected: the manpage calls this out as a limitation. Messages from diff or cmp can refer to a temporary file rather than the names you supplied. The wrapper removes that temporary file once the comparison finishes.

Keep the compressed originals until you have inspected an important result. This command is read-only towards those inputs, but a comparison only tells you their decompressed contents differ or match; it does not prove either archive is the file you expected.

4. Use the paired-file shortcut

With one argument, bzdiff applies a specific naming convention. Pass /work/report.txt and it compares that uncompressed file with /work/report.txt.bz2:

$ bzdiff -u /work/report.txt
$ printf '%s\n' "$?"
0

Handy when an archive and its extracted counterpart sit beside each other. The suffix is not guessed from an arbitrary compressed pathname the way some archive tools guess names: give the base, uncompressed name and make sure the matching .bz2 file exists.

A missing pair is an input error. Check both paths without changing anything:

$ ls -l /work/report.txt /work/report.txt.bz2
$ test -r /work/report.txt && test -r /work/report.txt.bz2
$ printf '%s\n' "$?"

Status 0 means both files were readable. If either path is absent or unreadable, fix the path or the permissions through your normal process. Do not grant broad permissions or run the whole comparison as root just to paper over a path mistake.

5. Choose bzcmp for byte-oriented comparison

bzcmp handles compressed files the same way but invokes cmp instead of diff. Reach for it when you want a compact equality test, or byte and character positions rather than a patch-like view:

$ bzcmp /archive/old.txt.bz2 /archive/new.txt.bz2
/archive/old.txt.bz2 /archive/new.txt.bz2 differ: byte 7, line 2
$ printf '%s\n' "$?"
1

The exact diagnostic depends on the installed cmp. As with bzdiff, status 0 means equal and status 1 means different. Options pass straight through to cmp or diff, so check those manpages before putting a less common option into a script: the wrapper does not translate a bzdiff option into a bzcmp one.

6. Put the result in a script

Branch on the exit status instead of parsing the output. This treats a difference as an intentional result, while reporting an input or decompression problem separately:

if bzdiff -q /path/to/old.txt.bz2 /path/to/new.txt.bz2; then
    printf '%s\n' 'files match'
else
    status=$?
    if [ "$status" -eq 1 ]; then
        printf '%s\n' 'files differ'
    else
        printf 'comparison failed with status %s\n' "$status" >&2
        exit "$status"
    fi
fi

-q is a diff option on systems where the installed diff supports it, and it suppresses the normal difference listing. If your script must work across several platforms, verify diff -q locally first. A non-zero result is not automatic proof of changed content either: a missing file, an unreadable input, invalid bzip2 data, or a downstream tool error all need investigating.

Common traps

• Do not unpack first out of habit. That creates extra data and can leave stale extracted copies that someone later mistakes for current content. Use bzdiff or bzcmp for the comparison, and write an extracted file only when a later tool genuinely needs one.
• Watch where you redirect. bzdiff writes differences to standard output, so > result.patch is safe when result.patch is a new destination, but shell redirection truncates an existing destination before bzdiff even starts. Choose a new name or copy the old result first.
• Exit status is not a monitoring alarm. The wrapper preserves the exit status from diff or cmp. A monitoring check that treats every non-zero status as an outage will misreport ordinary differences; capture the status immediately and decide whether status 1 is expected for your workflow.