No description, website, or topics provided.
C++ C Shell M4 Makefile
Latest commit 72e41af Jul 12, 2016 @jessek Add release hashes

This is md5deep, a set of cross-platform tools to compute hashes, or message digests, for any number of files while optionally recursively digging through the directory structure. It can also take a list of known hashes and display the filenames of input files whose hashes either do or do not match any of the known hashes. This version supports MD5, SHA-1, SHA-256, Tiger, and Whirlpool hashes.

See the file NEWS for a list of changes between releases.

See the file COPYING for information about the licensing for this program.

See the file INSTALL for (generic) compilation and installation instructions. Here's the short version that should just work in many cases:

sh # runs autoconf, automake
make install

Note that you must be normally root to install to the default location. The sudo command is helpful for doing so. You can specify an alternate installation location using the --prefix option to the configure script. For example, to install to /home/foo/bin, use:

$ ./configure --prefix=/home/foo

There is complete documentation on how to use the program on the project's homepage,

Release hashes

The latest release is version 4.4. Here are the hashes for the Win32 versions:

  • MD5 047ff5311d821f667150f37695bd01b3
  • SHA256 d5e85933e74e5ba6a73f67346bc2e765075d26949c831a428166c92772f67dbc

md5deep vs. hashdeep

For historical reasons, the program has different options and features when run with the names "hashdeep" and "md5deep."

hashdeep has a feature called "audit" which can use a list of known hashes to audit a set of FILES. Errors are reported to standard error. If no FILES are specified, reads from standard input.

-a Audit mode. Each input file is compared against the set of knowns. An
   audit is said to pass if each input file is matched against exactly
   one file in set of knowns. Any collisions, new files, or missing files
   will make the audit fail. Using this flag alone produces a message,
   either "Audit passed" or "Audit Failed".
   -v = prints the number of files in each category
   -v -v = prints all discrepancies
   -v -v -v = prints the results for every file examined and every known file.

-k <file> - The -k option must be used to load the audit file

To perform an audit:

hashdeep -r dir  > /tmp/auditfile            # Generate the audit file
hashdeep -a k /tmp/auditfile -r dir          # test the audit

Notice that the audit is performed with a standard hashdeep output file. (Internally, the audit is computed as part of the hashing process.)

Unicode Issues

POSIX-based modern computer systems consider filenames to be a sequence of bytes that are rendered as the application wishes. This means that filenames typically contain ASCII but can contain UTF-8, UTF-16, latin1, or even invalid Unicode codings.

Windows-based systems have one set of API calls for ASCII-based filenames and another set for filenames encoded as UCS-2, which "produces a fixed-length format by simply using the code point as the 16-bit code unit and produces exactly the same result as UTF-16 for 63,488 code points in the range 0-0xFFFF" according to [wikipedia] ( But wikipedia disputes the factual accuracy of this statement on the talk page. it's pretty clear that nobody is entirely sure that Windows actually does, and Windows itself may not be consistent.

Version 3 of this program addressed this issue by using the TCHAR variable to hold filenames on Windows and refusing to print them, priting a "?" instead. Version 4 of this program translates TCHAR strings to std::string strings at the soonest opportunity using the [Windows function WideCharToMultiByte] ( Flags have been added escape Unicode when it is printed.

There is no way (apparently) on Windows to open a UTF-8 filename; it needs to be converted back to a multi-byte filename with MultiByteToWideChar.

Fortunately, we never really need to convert back.

Notice that on Windows the files hashed can have unicode characters but the file with the hashes must have an ASCII name.


-D_UNICODE causes TCHAR to be defined as 'wchar_t'.


-D_UNICODE is not defined, causing TCHAR to be defined as 'char'.

Previously, win32 functions were controlled with #ifdef statements, like this:

#ifdef _WIN32
  if (NULL == realpath(fn,d_name))
    return TRUE;

There was also a file called tchar-local.h which actually changed the semantics of functions on different platforms, with things like this:

   #define  _tcsncpy   strncpy
   #define  _tstat_t   struct stat

This made the code very difficult to maintain.

With the 4.0 rewrite, we have changed this code with C++ functions that return objects were possible and avoid the use of #defines that so that on _WIN32 systems the function realpath() gets defined prior to its use, and the mainline code lacks the realpath() function. You can see this in cycles.cpp:

/* Return the canonicalized absolute pathname in UTF-8 on Windows and POSIX systems */
std::string get_realpath(const TCHAR *fn)
#ifdef _WIN32
     * expand a relative path to the full path.
    TCHAR absPath[PATH_MAX];
    if(_fullpath(absPath,fn,PAT_HMAX)==0) return "";
    return tchar_to_utf8(absPath);
    char resolved_name[PATH_MAX];	//
    if(realpath(fn,resolved_name)==0) return "";
    return string(resolved_name);

You can install mingw and then simply configure with something like this:

$ export PATH=$PATH:/usr/local/i386-mingw32-4.3.0/bin $ ./configure --host=i386-mingw32

Hash Algorithm References

The MD5 algorithm is defined in RFC 1321:

The SHA1 algorithm is defined in FIPS 180-1:

The SHA256 algorithm is defined FIPS 180-2:

The Tiger algorithm is defined at:

The Whirlpool algorithm is defined at:

Theory of Operation

  • main.cpp
    • sets up the system
  • dig.cpp
    • iterates through the individual directories
    • calls hash_file() in hash.cpp for each file to hash
  • hash.cpp
    • performs the hashing of each file
  • display.cpp
    • stores/displays the results