Re: New Defects reported by Coverity Scan for zsh

[Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx>]

Why am I getting this?  Who set this up?

On Tue, Feb 28, 2023 at 5:01 AM <scan-admin@xxxxxxxxxxxx> wrote:
>
> >>>     CID 1521554:  Control flow issues  (MISSING_RESTORE)
> >>>     Value of non-local "*ss" that was saved in "sav" is not restored as it was along other paths.
> 2159                    return NULL;

Pointer to heap memory, not used again, no need to restore.

> /Src/params.c: 6268 in upscope()
> >>>     Null-checking "pm" suggests that it may be null, but it has already been dereferenced on all paths leading to the check.

Caller shouldn't ever pass NULL.  Is this going to keep complaining about it?

> *** CID 1521548:  Memory - illegal accesses  (USE_AFTER_FREE)
> /Src/builtin.c: 1211 in cd_new_pwd()
> 1205            zsfree(getlinknode(dirstack));
> 1206
> 1207         if (chasinglinks) {
> 1208            s = findpwd(new_pwd);
> 1209            if (s) {
> 1210                zsfree(new_pwd);
> >>>     CID 1521548:  Memory - illegal accesses  (USE_AFTER_FREE)
> >>>     Using freed pointer "s".
> 1211                new_pwd = s;
> 1212            }

This is a knock-on to the complaint about findpwd() below.

> 7181            if (meta) {
> >>>     CID 1521546:  Uninitialized variables  (UNINIT)
> >>>     Using uninitialized value "t[-1]".
> 7182                t[-1] |= 0x80;
> 7183                meta = 0;
> 7184            }

Hm, I guess "t" might not have advanced past its original starting
assignment if control passes through the #ifdef MULTIBYTE block about
60 lines earlier, without returning?

#ifdef MULTIBYTE_SUPPORT
        } else if ((how & GETKEY_SINGLE_CHAR) &&
                   isset(MULTIBYTE) && (unsigned char) *s > 127) {
            wint_t wc;
            int len;
            len = mb_metacharlenconv(s, &wc);
            if (wc != WEOF) {
                *misc = (int)wc;
                return s + len;
            }
#endif


> *** CID 1521545:  Resource leaks  (RESOURCE_LEAK)
> /Src/Modules/param_private.c: 130 in makeprivate()
> >>>     CID 1521545:  Resource leaks  (RESOURCE_LEAK)
> >>>     Variable "gsu" going out of scope leaks the storage it points to.

Can't happen unless the definition of PM_TYPE() changes without this
code being updated.

> *** CID 1521544:  Memory - illegal accesses  (USE_AFTER_FREE)
> /Src/utils.c: 801 in findpwd()
> 795
> 796         if (*s == '/')
> 797             return xsymlink(s, 0);
> 798         s = tricat((pwd[1]) ? pwd : "", "/", s);
> 799         t = xsymlink(s, 0);
> 800         zsfree(s);
> >>>     CID 1521544:  Memory - illegal accesses  (USE_AFTER_FREE)
> >>>     Using freed pointer "t".
> 801         return t;
> 802     }

Not seeing how it calculates this one, I think xsymlink(s,0) is going
to end up returning either a pointer to the static mbuf[] in metafy(),
or heap memory.  Anyone else see an alternative?  Is it treating mbuf
as freed stack even though it is declared static?