Fixed salt generation when counter == 0 for LessPass compatibility

[myau-def]

Fixes the bug where counter = 0 produces an incorrect salt.

Description

There is a bug in salt generation when counter is set to 0. It breaks compatibility with the official LessPass implementation.

For any counter >= 1, the generated passwords match perfectly. However, for counter = 0, the generated passwords diverge.

Steps to Reproduce

With counter = 0 (Bug):

$ lesspass "site" "login" "password" -C 0
pkwuO9<-W">g,Zag

$ lesspass-rs "site" "login" "password" -c 0
7q~o>vDo6n(Rf+"g

With counter >= 1 (Works correctly):

$ lesspass "site" "login" "password" -C 1
|KTnP2D^64'`q]Er

$ lesspass-rs "site" "login" "password" -c 1
|KTnP2D^64'`q]Er

$ lesspass "site" "login" "password" -C 2
b/f@D|kl2Rs:9k\^

$ lesspass-rs "site" "login" "password" -c 2
b/f@D|kl2Rs:9k\^

Root Cause

The issue is located in the generate_salt_to_uninit function (line 138). When counter is 0, the while counter != 0 loop condition is immediately false, so the loop never executes. As a result, it returns an empty buffer instead of "0".

let mut counter_buf = [MaybeUninit::uninit(); 8];
let counter = {
    let mut counter = counter as usize;
    let mut i = counter_buf.len();
    
    // This works for counter >= 1, but fails for counter = 0
    while counter != 0 {
        counter_buf[i - 1].write(b"0123456789abcdef"[counter & 0xf]);
        counter >>= 4;
        i -= 1;
    }

    &counter_buf[i..] // Returns an empty slice &[] when counter is 0
};

[71]

Thank you and sorry for the slow response!

Could you add 0 in the test cases here (and regenerate the .rs test file), so the behavior is tested from now on:

lesspass.rs/tests/make-blackbox-tests.sh

Line 14 in 7befe4b

for COUNTER in 1 8 32 100000; do

[myau-def]

Done! I've updated the make-blackbox-tests.sh script to include 0 in the COUNTER loops.

While I was at it, I also noticed that the generator script was using legacy CLI flags (--no-lowercase instead of --no-lower, etc.) which caused panic messages to bleed into the test vector output. I patched the bash script to properly map those arguments and updated the boilerplate template so that tests/blackbox.rs builds and tests successfully out of the box.

The test file has been fully regenerated and the entire history was cleaned up via a force-push.

[71]

Thank you and sorry for the slow response again. I will squash before merging anyway, so please push new commits instead of rebasing so it's easy to see a diff between revisions.

[71]

nit: if bash is needed, I'd rather use

Suggested change

	#!/bin/bash
	#!/usr/bin/env -S bash -euo pipefail

I used /bin/sh to remove dependencies and keep things simple, but if that's not an option, I prefer the stricter / env-dependent shebang.

[71]

Can you remove the non-English comments, please?

[71]

To avoid the large diff, could you format this file please? And possibly add cargo fmt or something like that to the end of make-blackbox-tests.sh to make sure this is done automatically.