337017 - Coverity 521 NEGATIVE_RETURNS in NSS_SMIMEUtil_FindBulkAlgForRecipients

Status: RESOLVED INVALID

(NSS :: Libraries, enhancement, P4)

Description

[timeless]

I know the code implies that this can't happen, but i'd much rather we handle it.

Event negative_return_fn: Called negative-returning function "smime_mapi_by_cipher" [model]
Event var_assign: NEGATIVE return value of "smime_mapi_by_cipher" assigned to signed variable "mapi"
Also see events: [var_assign][negative_returns]

561  	    mapi = smime_mapi_by_cipher(cipher);
562  	

At conditional (1): "i < smime_cipher_map_count" taking true path
At conditional (3): "i < smime_cipher_map_count" taking true path
At conditional (5): "i < smime_cipher_map_count" taking true path
At conditional (7): "i < smime_cipher_map_count" taking false path

164  	    for (i = 0; i < smime_cipher_map_count; i++) {

At conditional (2): "(smime_cipher_map[i]).cipher == cipher" taking false path
At conditional (4): "(smime_cipher_map[i]).cipher == cipher" taking false path
At conditional (6): "(smime_cipher_map[i]).cipher == cipher" taking false path

165  		if (smime_cipher_map[i].cipher == cipher)
166  		    return i;	/* bingo */
167  	    }

Event return_negative_constant: Explicitly returning NEGATIVE value "-1"

168  	    return -1;		/* should not happen if we're consistent, right? */

Event negative_returns: Tracked variable "mapi" used as index to array "smime_cipher_map"
Also see events: [negative_return_fn][var_assign]

563  	    *bulkalgtag = smime_cipher_map[mapi].algtag;

Comment 1

[Nelson Bolyard (seldom reads bugmail)]

There is no bug here.  

563          cipher = smime_choose_cipher(NULL, rcerts);
564          mapi = smime_mapi_by_cipher(cipher);
565
566          *bulkalgtag = smime_cipher_map[mapi].algtag;

Function smime_choose_cipher ALWAYS returns a cipher that is in the mapping 
table.  Function smime_mapi_by_cipher will ALWAYS return a valid index into
that mapping table if it is given a cipher in that table.  So, the result 
is that mapi cannot have an out-of-bounds value.  

Now, coverity guys, is there some comment we can put into the code here (like the comments one puts into to silence lint) that will silence coverity about this?

Comment 2

[Nelson Bolyard (seldom reads bugmail)]

I marked this as a FALSE POSITIVE in coverity.
Reopen this bug if coverity won't take "no" (FALSE) for an answer.