bignum_from_string: only works with hex-strings w/ multiple of 8 characters

[DEF7]

The problem lies with this piece of code:

tiny-bignum-c/bn.c

Lines 108 to 120 in e814d2b

	int i = nbytes - (2 * WORD_SIZE); /* index into string */
	int j = 0; /* index into array */
	/* reading last hex-byte "MSB" from string first -> big endian */
	/* MSB ~= most significant byte / block ? :) */
	while (i >= 0)
	{
	tmp = 0;
	sscanf(&str[i], SSCANF_FORMAT_STR, &tmp);
	n->array[j] = tmp;
	i -= (2 * WORD_SIZE); /* step WORD_SIZE hex-byte(s) back in the string. */
	j += 1; /* step one element forward in the array. */
	}

which appears as:

`
int i = nbytes - (2 * WORD_SIZE); /* index into string /
int j = 0; / index into array */

/* reading last hex-byte "MSB" from string first -> big endian /
/ MSB ~= most significant byte / block ? :) /
while (i >= 0)
{
tmp = 0;
sscanf(&str[i], SSCANF_FORMAT_STR, &tmp);
n->array[j] = tmp;
i -= (2 * WORD_SIZE); / step WORD_SIZE hex-byte(s) back in the string. /
j += 1; / step one element forward in the array. */
}
`

If 'i' (the length of the inbound string, in characters) is not a multiple of '2 * WORD_SIZE' then it will skip over the MSB hexadecimal characters at the beginning of the string when 'i < 0' and miss the characters at the beginning of the inbound string, incorrectly setting 'n'.

I think the use of sprintf/sscanf is not ideal. Custom to/from string routines would be better.

[kokke]

Hi @DEF7 and thanks for your interest in this project :)

You're correct that the to/from_string methods only works for a 8-bite chunks, but this is by design.

I only use the functions for the test-code, which I hint at in the readme: **stdio.h is only used for testing functions parsing to and from hex-strings.**. This is the reason for not implementing better to/from_string methods -- they are only used for the testing.

Custom to/from string routines would be better.

I agree. I would most likely accept a PR that improved the current situation :)

[DEF7]

Ok. In the meantime I'll just pad the beginning of my hex strings with zeroes which seems to work fine for now. Thanks!

…

On Mon, Jan 20, 2020 at 8:09 AM kokke ***@***.***> wrote: Hi @DEF7 <https://github.com/DEF7> and thanks for your interest in this project :) You're correct that the to/from_string methods only works for a 8-bite chunks, but this is by design. I only use the functions for the test-code, which I hint at in the readme: **stdio.h is only used for testing functions parsing to and from hex-strings.**. This is the reason for not implementing better to/from_string methods -- they are only used for the testing. Custom to/from string routines would be better. I agree. I would most likely accept a PR that improved the current situation :) — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#14?email_source=notifications&email_token=ACYGB5KE4LJCCWBACS4HB73Q6XEDJA5CNFSM4KJGGWJ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEJND4ZQ#issuecomment-576339558>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACYGB5IC2GGINBAONB6TDOTQ6XEDJANCNFSM4KJGGWJQ> .

[kokke]

Until further notice, I think it's fair to add this piece of assertion-code in bignum_from_string:

require((nbytes % (sizeof(DTYPE) * 2)) == 0, "string length must be a multiple of (sizeof(DTYPE) * 2) characters");

,

... so that you at least get a proper failure/crash with a line number etc., if you mess up:

,

test_hand_picked: bn.c:104: bignum_from_string: Assertion `(nbytes % (sizeof(uint16_t) * 2)) == 0 && "string length must be a multiple of (sizeof(DTYPE) * 2) characters"' failed.

What do you think @DEF7 ?