Solar Designer

> After an easy change in `valid()` it can detect and complain. This comment applied only to the specific issue I had referenced, but there are now many more CPU-only...

In the RAR code, these are false positives because the array accesses are right after checks of `i < sizeof(bit_length) / sizeof(bit_length[0])`. I don't know what's the best way to...

Looks like this can provide great speedup for S-box lookups over bitslice: 22.125 VPTERNLOG's per 512 lookups vs. 1 VPERMB per 64, so kind of a ~2.76x speedup for this...

Looks like this won't help for DES, but it might for Lotus where we need the 8-bit outputs and don't need to expand them further, and where the S-box expressions...

> each 4-bit output of S-box lookup then goes into `P`, which we'd need to implement by separating the 4 bits into up to 4 different vectors, so that they're...

> I belive for multipart files you'd need to concatenate them and then run rar2john on the resulting single file. If so, maybe `rar2john` should refuse to run on the...

> We also need to recall why it ended up 125 rather than 127 now. Do we possibly use this extra room somewhere? I found no intentional uses of the...

This hack helps: ```diff +++ b/src/external.c @@ -197,18 +197,18 @@ int ext_has_function(const char *mode, const char *function) void ext_init(char *mode, struct db_main *db) { ext_minlen = options.eff_minlength; - maxlen =...

Also, @jfoug's hybrid_example was totally broken in several ways. ```diff @@ -4181,23 +4183,23 @@ void new() * word will be too long to be used, or too short to be...

> FWIW, I find this stuff by running with greatly reduced maximums, not increased at all, and a build with ASan. For such testing, I have to temporarily remove this:...