On 2023-09-03 19:58, Celeste wrote:
> Thanks for taking the time to look into this issue, Kaz.
> 
> The patch, however, does not solve the failing test on musl.
> 
> (crypt "a" "*$") returns "*$OGH2VZh3lDE", which is also the output of `mkpasswd -m des -S '*$' a`, in case you want to try it in an Alpine VM.

OK, thanks, I have a better idea now.

Let's make the test 

   (crypt "a" "::") :error

Musl's crypt is actually routing all the unrecognized salt codes to DES.
There is no $9$ algorithm.

Its DES implementation accepts salts that are invalid in traditional
crypt; it accepts almost anything.

According to the code, only these characters are disallowed:
NUL, NL and : (colon):

  /*
   * When we choose to "support" invalid salts, nevertheless disallow those
   * containing characters that would violate the passwd file format.
   */
  static inline int ascii_is_unsafe(unsigned char ch)
  {
	return !ch || ch == '\n' || ch == ':';
  }

My motivation is that I would like to have at least one test which elicits
a null return from crypt due to bad input, in addition to the
(crypt "a" "b") :error test.