Re: [PATCH] scripts/kconfig/menu.c warning for uninitialized "jump"

[Michal Marek <mmarek@suse.cz>]

Dne 28.11.2014 v 19:50 Joe Perches napsal(a):
> On Fri, 2014-11-28 at 14:33 +0100, Michal Marek wrote:
>> On 2014-11-21 05:22, Peter Teoh wrote:
>>> This warning was found in v3.18-rc3-68-g20f3963 of Linus git-tree.
>>>
>>>   SHIPPED scripts/kconfig/zconf.hash.c
>>>   HOSTCC  scripts/kconfig/zconf.tab.o
>>> In file included from scripts/kconfig/zconf.tab.c:2537:0:
>>> scripts/kconfig/menu.c: In function ‘get_symbol_str’:
>>> scripts/kconfig/menu.c:590:18: warning: ‘jump’ may be used
>>> uninitialized in this function [-Wmaybe-uninitialized]
>>>      jump->offset = strlen(r->s);
>>>                   ^
>>> scripts/kconfig/menu.c:551:19: note: ‘jump’ was declared here
>>>   struct jump_key *jump;
>>>                    ^
>>
>> First of all, the warning is bogus (the condition under which 'jump' is
>> used is stronger than that under which 'jump' is initialized). But since
>> people have been reporting the warning on and off for some time, we have
>> to shut it up somehow, as the affected gcc versions are not dying out,
>> apparently.
>>
>>
>>> -                       if (head && location && menu == location)
>>> +                       if (head && location && (menu == location) && (jump))
>>>                                 jump->offset = strlen(r->s);
>>
>> Let's assume, for the sake of argument, that gcc is right and jump may
>> be uninitialized here. Then the added check for jump being non-null just
>> tests an uninitialized variable and thus behaves randomly. It prevents
>> the code from writing to NULL->offset, but does not prevent it from
>> writing to <random address>->offset.
> 
> Maybe the 'right' thing to do is to mark the xmalloc
> prototype as "__attribute__((returns_nonnull))"

The warning says that 'jump' can be used uninitialized (random), not
that it can be NULL.

Michal