bsdiff and bspatch are libraries for building and applying patches to binary files.
The original algorithm and implementation was developed by Colin Percival. The algorithm is detailed in his (unpublished) paper, Naïve Differences of Executable Code. For more information, visit his website at http://www.daemonology.net/bsdiff/.
I maintain this project seperately from Colin's work, with the goal of making the core functionality easily embedable in existing projects.
@MatthewEndsley
https://github.com/mendsley/bsdiff
Copyright 2003-2005 Colin Percival
Copyright 2012 Matthew Endsley
This project is governed by the BSD 2-clause license. For details see the file titled LICENSE in the project root folder.
There are two separate libraries in the project, bsdiff and bspatch. Each are self contained in bsdiff.c and bspatch.c The easiest way to integrate is to simply copy the c file to your source folder and build it.
The overarching goal was to modify the original bsdiff/bspatch code from Colin and eliminate external dependencies and provide a simple interface to the core functionality.
I've exposed relevant functions via the _stream
classes. The only external
dependency not exposed is memcmp
in bsdiff
.
This library generates patches that are not compatible with the original bsdiff tool. The impompatibilities were motivated by the patching needs for the game AirMech https://www.carbongames.com and the following requirements:
- Eliminate/minimize any seek operations when applying patches
- Eliminate any required disk I/O and support embedded streams
- Ability to easily embed the routines as a library instead of an external binary
- Compile+run on all platforms we use to build the game (Windows, Linux, NaCl, OSX)
The libraries should compile warning free in any moderately recent version of
gcc. The project uses <stdint.h>
which is technically a C99 file and not
available in Microsoft Visual Studio. The easiest solution here is to use the
msinttypes version of stdint.h from https://code.google.com/p/msinttypes/.
The direct link for the lazy people is:
https://msinttypes.googlecode.com/svn/trunk/stdint.h.
If your compiler does not provide an implementation of <stdint.h>
you can
remove the header from the bsdiff/bspatch files and provide your own typedefs
for the following symbols: uint8_t
, uint64_t
and int64_t
.
Each project has an optional main function that serves as an example for using
the library. Simply defined BSDIFF_EXECUTABLE
or BSPATCH_EXECUTABLE
to
enable building the standalone tools.
struct bsdiff_stream
{
void* opaque;
void* (*malloc)(size_t size);
void (*free)(void* ptr);
int (*write)(struct bsdiff_stream* stream,
const void* buffer, int size);
};
int bsdiff(const uint8_t* old, int64_t oldsize, const uint8_t* new,
int64_t newsize, struct bsdiff_stream* stream);
In order to use bsdiff
, you need to define functions for allocating memory and
writing binary data. This behavior is controlled by the stream
parameted
passed to to bsdiff(...)
.
The opaque
field is never read or modified from within the bsdiff
function.
The caller can use this field to store custom state data needed for the callback
functions.
The malloc
and free
members should point to functions that behave like the
standard malloc
and free
C functions.
The write
function is called by bsdiff to write a block of binary data to the
stream. The return value for write
should be 0
on success and non-zero if
the callback failed to write all data. In the default example, bzip2 is used to
compress output data.
bsdiff
returns 0
on success and -1
on failure.
struct bspatch_stream
{
void* opaque;
int (*read)(const struct bspatch_stream* stream,
void* buffer, int length);
};
int bspatch(const uint8_t* old, int64_t oldsize, uint8_t* new,
int64_t newsize, struct bspatch_stream* stream);
The bspatch
function transforms the data for a file using data generated from
bsdiff
. The caller takes care of loading the old file and allocating space for
new file data. The stream
parameter controls the process for reading binary
patch data.
The opaque
field is never read or modified from within the bspatch function.
The caller can use this field to store custom state data needed for the read
function.
The read
function is called by bspatch
to read a block of binary data from
the stream. The return value for read
should be 0
on success and non-zero
if the callback failed to read the requested amount of data. In the default
example, bzip2 is used to decompress input data.
bspatch
returns 0
on success and -1
on failure. On success, new
contains
the data for the patched file.