ext/hash: Fix GH-16711: Segfault in mhash()

[Girgias]

Rewrote part of how it works to handle all cases dealing with the mhash_to_hash struct.

[cmb69]

From a quick look, the only affected function is mhash(). Can't we just bail out if algo == NULL after

php-src/ext/hash/hash.c

Lines 1211 to 1217 in fde053b

	/* need to convert the first parameter from int constant to string algorithm name */
	if (algorithm >= 0 && algorithm < MHASH_NUM_ALGOS) {
	struct mhash_bc_entry algorithm_lookup = mhash_to_hash[algorithm];
	if (algorithm_lookup.hash_name) {
	algo = zend_string_init(algorithm_lookup.hash_name, strlen(algorithm_lookup.hash_name), 0);
	}
	}

[nielsdos]

I agree with @cmb69. This patch is too complex for a bugfix as it adds more sources of truth (we have to check the algorithm ID and name separately and assert they're consistent).

[cmb69]

IMO okay now. Could also raise a warning or throw an exception – I don't mind, since these functions are deprecated anyway.

[Girgias]

IMO okay now. Could also raise a warning or throw an exception – I don't mind, since these functions are deprecated anyway.

Meh, they return false on error with no warnings, introducing one now seems kinda pointless. Also not sure how many people even use mhash.

[nielsdos]

Also not sure how many people even use mhash.

It's deprecated anyway, it doesn't make much sense to do more than the bare minimum keeping it working IMHO.

[nielsdos]

LGTM thanks!