Open solardiz opened 2 years ago
Now that we have #4728, we might want to simply drop the hard-coded check(s) against last
(and copying to last
or swapping of two buffer pointers). On very fast hashes, these checks were possibly unjustified overhead, and on slower hashes they're redundant with the new suppressor.
A long-standing core bug: the wordlist/rules check against
last
word (poor man's de-duplication) isn't doing the right thing when an external mode filter modifies the word (the check compares post-filter vs. pre-filter). This appears non-trivial to fix while keeping the currentlast
logic with two buffers coming fromrules.c
. Maybe something to avoid in a rewrite of that logic.