Imported self-test enhancements from 1.7.6-jumbo-12. |
|
More...
|
almost 14 years ago
|
Increased SHADOW_HASH_LOG from 8 to 18. Combined with the recent change to login_hash(), this provides 10x speedup for unshadow on 100k entry test files. |
|
More...
|
almost 14 years ago
|
Introduced some parallelism into login_hash() similarly to the way it was done in unique.c: line_hash(). Made it work for larger hash tables and process longer login names (without discarding any characters from the computation). |
|
More...
|
almost 14 years ago
|
A further optimization to line_hash(), approx 2% speedup overall on a certain test case. |
|
More...
|
almost 14 years ago
|
In init_hash(), use memset() to initialize buffer.hash[] with ENTRY_END_HASH. |
|
More...
|
almost 14 years ago
|
Use fwrite() instead of fprintf(). Combined with the change to line_hash(), this provides a speedup of 12% to 25% over JtR 1.7.6's revision of unique.c on different realistic test cases. |
|
More...
|
almost 14 years ago
|
Introduced some parallelism into line_hash() similarly to the way it was done for single_key_hash(). This provides a 7% overall speedup (for the entire program invocation) on a certain realistic test case. |
|
More...
|
almost 14 years ago
|
When computing hashes of hash encoding strings, start from the second to last character, not from the very last one, because the latter contains fewer bits of the actual hash value with many hash types. |
|
More...
|
almost 14 years ago
|
Stagger the crypt_data structs to reduce their competition for the same cache lines. |
|
More...
|
almost 14 years ago
|
Expanded the copyright years. |
|
More...
|
almost 14 years ago
|
Require gcc for AVX because DES_bs_all is currently being aligned in a gcc-specific way. |
|
More...
|
almost 14 years ago
|
Use mem_alloc_tiny(..., MEM_ALIGN_PAGE) rather than calloc() to allocate instances of "struct crypt_data" in c3_fmt.c. |
|
More...
|
almost 14 years ago
|
Added Intel AVX and AMD XOP instruction sets support for bitslice DES (with C compiler intrinsics). |
|
More...
|
almost 14 years ago
|
In tty_init(), don't do anything at all when compiled with DJGPP (for DOS). JtR 1.7.6 erroneously introduced some no-op code for that case. |
|
More...
|
over 14 years ago
|
In the description of "--format", only list "crypt" if actually supported in the current build. |
|
More...
|
over 14 years ago
|
1.7.6.1 |
|
More...
|
over 14 years ago
|
Corrected a logic error introduced in JtR 1.7.4.2: in "single crack" mode, we need a salt's key buffer even when we have no words corresponding to that salt's hashes to base candidate passwords on. We need this buffer to hold other salts' successful guesses for testing against this salt's hashes. |
|
More...
|
over 14 years ago
|
1.7.6 |
|
More...
|
over 14 years ago
|
Call tty_done() when forcibly terminating John from a signal handler. tty_done() is mostly async signal safe (the worst that can happen is that we restore the tty mode and attempt to close the fd twice). |
|
More...
|
over 14 years ago
|
If we're in "--stdin" mode (reading candidate passwords from stdin), then only initialize the tty if stdin is not a tty. Otherwise it could be the same tty, in which case we'd interfere with the user's ability to type candidate passwords directly to John. |
|
More...
|
over 14 years ago
|
Detect whether the hashes being loaded are supported by the system or not (and avoid loading unsupported ones). |
|
More...
|
over 14 years ago
|
Take into consideration that SHA-crypt hash encoding strings may be shorter when the salts are shorter than usual or missing. |
|
More...
|
over 14 years ago
|
Revised the detection of salts. |
|
More...
|
over 14 years ago
|
When loading username-less hashes, don't enter the question marks (which are to be displayed by John in place of the missing usernames) for processing by the "single crack" mode (this was a side-effect of the initial implementation). |
|
More...
|
over 14 years ago
|
Extended the generic crypt(3) support to Solaris, including optional OpenMP parallelization and detection of matching salts with SunMD5 hashes. |
|
More...
|
over 14 years ago
|
Use "__inline__" rather than "inline" with pre-3.1 GCC, which matters when compiling with non-default options (e.g., "-pedantic"). (On GCC 3.1+, we use __attribute__((always_inline)) instead.) |
|
More...
|
over 14 years ago
|
Updates on supported hash types. |
|
More...
|
over 14 years ago
|
Only state that John "normally" does not use a crypt(3)-style routine, not that it does not use one at all, which is no longer the case with the newly-introduced generic crypt(3) support. |
|
More...
|
over 14 years ago
|
#include <stdlib.h> for calloc(3) |
|
More...
|
over 14 years ago
|
Updates related to DES S-box expressions. |
|
More...
|
over 14 years ago
|