'\0\' is not correctly handled

[triska]

For a test case, please create a file with two bytes:

$ echo -n -e "\x0\x1" > f

Confirm the content (two bytes, in hex: 00 01) with:

$ hexdump f
0000000 00 01
0000002

However, the following unexpectedly succeeds:

:- use_module(library(pio)).
   true.
?- phrase_from_file([L], f).
   L = '\x1\'.

Expected: Failure, on the grounds that the file contains 2 bytes.

For comparison, the following works as intended:

$ echo -n -e "\x1\x2" > f

yielding:

?- phrase_from_file([A,B], f).
   A = '\x1\', B = '\x2\'.

[triska]

Here is a shorter test case:

?- Xs = "\x0\".
   Xs = [].

Expected: Xs = "\x0\".

[ghost]

This issue is linked to this one.

[triska]

Ah yes, this is already filed, thank you a lot!

I'm closing this issue.

Normally, zero-bytes are very rare, but a major problem is that they easily arise during encryption. For instance, the following query should never succeed (since the length of the cipher text does not match the plain text), but now actually succeeds:

?- repeat,
   crypto_n_random_bytes(32, Key),
   crypto_n_random_bytes(12, IV),
   PlainText = "a",
   crypto_data_encrypt(PlainText, 'chacha20-poly1305', Key, IV, CipherText, []),
   \+ same_length(PlainText, CipherText).

This means that we cannot reliably use the symmetric encryption facilities as long as this issue is present.