[PATCH resend] kconfig: Fix compiler warning in menu.c

[PATCH resend] kconfig: Fix compiler warning in menu.c

[Hans de Goede]

Hi,

This one seems to have fallen through the cracks.

Regards,

Hans

[U-Boot] [PATCH resend] kconfig: Fix compiler warning in menu.c

[Hans de Goede]

Hi,

This one seems to have fallen through the cracks.

Regards,

Hans

[PATCH resend] kconfig: Fix compiler warning in menu.c

[Hans de Goede]

This fixes the following compiler warning:

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);
                  ^
In file included from scripts/kconfig/zconf.tab.c:2537:0:
scripts/kconfig/menu.c:551:19: note: ‘jump’ was declared here
  struct jump_key *jump;

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
---
 scripts/kconfig/menu.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/scripts/kconfig/menu.c b/scripts/kconfig/menu.c
index a26cc5d..584e0fc 100644
--- a/scripts/kconfig/menu.c
+++ b/scripts/kconfig/menu.c
@@ -548,7 +548,7 @@ static void get_prompt_str(struct gstr *r, struct property *prop,
 {
 	int i, j;
 	struct menu *submenu[8], *menu, *location = NULL;
-	struct jump_key *jump;
+	struct jump_key *jump = NULL;
 
 	str_printf(r, _("Prompt: %s\n"), _(prop->text));
 	menu = prop->menu->parent;
-- 
2.1.0

[U-Boot] [PATCH resend] kconfig: Fix compiler warning in menu.c

[Hans de Goede]

This fixes the following compiler warning:

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);
                  ^
In file included from scripts/kconfig/zconf.tab.c:2537:0:
scripts/kconfig/menu.c:551:19: note: ?jump? was declared here
  struct jump_key *jump;

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
---
 scripts/kconfig/menu.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/scripts/kconfig/menu.c b/scripts/kconfig/menu.c
index a26cc5d..584e0fc 100644
--- a/scripts/kconfig/menu.c
+++ b/scripts/kconfig/menu.c
@@ -548,7 +548,7 @@ static void get_prompt_str(struct gstr *r, struct property *prop,
 {
 	int i, j;
 	struct menu *submenu[8], *menu, *location = NULL;
-	struct jump_key *jump;
+	struct jump_key *jump = NULL;
 
 	str_printf(r, _("Prompt: %s\n"), _(prop->text));
 	menu = prop->menu->parent;
-- 
2.1.0

Re: [U-Boot] [PATCH resend] kconfig: Fix compiler warning in menu.c

[Jeroen Hofstee]

Hello Hans,

On 12-10-14 12:25, Hans de Goede wrote:
> Hi,
>
> This one seems to have fallen through the cracks.
>
> Regards,
>
> Hans
>
(for U-boot)

nope, you replace an innocent warning (_might_ be) with
bad code, without any comment it is just because gcc failed
to recognize it is fine. Nor did you respond to the suggestion
if it helps gcc to recognize that if the two booleans are merged
into a single one. [or even split it in an if () if ()]. With this patch
you prevent any serious warning in case the variable is actually
used but not initialized, which is even worse if you ask me.

Regards,
Jeroen

[U-Boot] [PATCH resend] kconfig: Fix compiler warning in menu.c

[Jeroen Hofstee]

Hello Hans,

On 12-10-14 12:25, Hans de Goede wrote:
> Hi,
>
> This one seems to have fallen through the cracks.
>
> Regards,
>
> Hans
>
(for U-boot)

nope, you replace an innocent warning (_might_ be) with
bad code, without any comment it is just because gcc failed
to recognize it is fine. Nor did you respond to the suggestion
if it helps gcc to recognize that if the two booleans are merged
into a single one. [or even split it in an if () if ()]. With this patch
you prevent any serious warning in case the variable is actually
used but not initialized, which is even worse if you ask me.

Regards,
Jeroen

[U-Boot] [PATCH resend] kconfig: Fix compiler warning in menu.c

[Simon Glass]

Hi Jeroen,

On 12 October 2014 10:13, Jeroen Hofstee <jeroen@myspectrum.nl> wrote:

> Hello Hans,
>
> On 12-10-14 12:25, Hans de Goede wrote:
>
>> Hi,
>>
>> This one seems to have fallen through the cracks.
>>
>> Regards,
>>
>> Hans
>>
>>  (for U-boot)
>
> nope, you replace an innocent warning (_might_ be) with
> bad code, without any comment it is just because gcc failed
> to recognize it is fine. Nor did you respond to the suggestion
> if it helps gcc to recognize that if the two booleans are merged
> into a single one. [or even split it in an if () if ()]. With this patch
> you prevent any serious warning in case the variable is actually
> used but not initialized, which is even worse if you ask me.
>

That is a pretty acerbic tone to take on the U-Boot list at least. Are you
two drinking buddies or something?

Regards,
Simon

Re: [U-Boot] [PATCH resend] kconfig: Fix compiler warning in menu.c

[Jeroen Hofstee]

Hello Simon,

On 13-10-14 07:14, Simon Glass wrote:
> Hi Jeroen,
>
> On 12 October 2014 10:13, Jeroen Hofstee <jeroen@myspectrum.nl> wrote:
>
>> Hello Hans,
>>
>> On 12-10-14 12:25, Hans de Goede wrote:
>>
>>> Hi,
>>>
>>> This one seems to have fallen through the cracks.
>>>
>>> Regards,
>>>
>>> Hans
>>>
>>>   (for U-boot)
>> nope, you replace an innocent warning (_might_ be) with
>> bad code, without any comment it is just because gcc failed
>> to recognize it is fine. Nor did you respond to the suggestion
>> if it helps gcc to recognize that if the two booleans are merged
>> into a single one. [or even split it in an if () if ()]. With this patch
>> you prevent any serious warning in case the variable is actually
>> used but not initialized, which is even worse if you ask me.
>>
> That is a pretty acerbic tone to take on the U-Boot list at least. Are you
> two drinking buddies or something?

no, it is because we have discussed this patch before and resending
it won't address the issue raised. But you are right, it is likely done with
less evil intends then I took it for, so let me explain my concern again
in a politer way. The problem is that gcc 4.9 starts warning in the
following case:

int *ptr;

if (a)
     ptr = something;

if (a && b)
     ptr->bla = value;
else
    do_something_else();


it will warn that ptr _might_ be used uninitialized (but it always is).
This is fixed in this patch by assigning NULL to ptr, and while that makes
the warning go away it actually prevents the valid warning, ptr _is_ used
uninitialized if you start using it in the else case. Hence my request if we
can't find a better solution for this.

Does anyone know a better solution for this or should we consider
disabling the might be unused warning?

Regards,
Jeroen

[U-Boot] [PATCH resend] kconfig: Fix compiler warning in menu.c

[Jeroen Hofstee]

Hello Simon,

On 13-10-14 07:14, Simon Glass wrote:
> Hi Jeroen,
>
> On 12 October 2014 10:13, Jeroen Hofstee <jeroen@myspectrum.nl> wrote:
>
>> Hello Hans,
>>
>> On 12-10-14 12:25, Hans de Goede wrote:
>>
>>> Hi,
>>>
>>> This one seems to have fallen through the cracks.
>>>
>>> Regards,
>>>
>>> Hans
>>>
>>>   (for U-boot)
>> nope, you replace an innocent warning (_might_ be) with
>> bad code, without any comment it is just because gcc failed
>> to recognize it is fine. Nor did you respond to the suggestion
>> if it helps gcc to recognize that if the two booleans are merged
>> into a single one. [or even split it in an if () if ()]. With this patch
>> you prevent any serious warning in case the variable is actually
>> used but not initialized, which is even worse if you ask me.
>>
> That is a pretty acerbic tone to take on the U-Boot list at least. Are you
> two drinking buddies or something?

no, it is because we have discussed this patch before and resending
it won't address the issue raised. But you are right, it is likely done with
less evil intends then I took it for, so let me explain my concern again
in a politer way. The problem is that gcc 4.9 starts warning in the
following case:

int *ptr;

if (a)
     ptr = something;

if (a && b)
     ptr->bla = value;
else
    do_something_else();


it will warn that ptr _might_ be used uninitialized (but it always is).
This is fixed in this patch by assigning NULL to ptr, and while that makes
the warning go away it actually prevents the valid warning, ptr _is_ used
uninitialized if you start using it in the else case. Hence my request if we
can't find a better solution for this.

Does anyone know a better solution for this or should we consider
disabling the might be unused warning?

Regards,
Jeroen

Re: [U-Boot] [PATCH resend] kconfig: Fix compiler warning in menu.c

[Tom Rini]

[-- Attachment #1: Type: text/plain, Size: 2405 bytes --]

On Mon, Oct 13, 2014 at 08:48:39AM +0200, Jeroen Hofstee wrote:
> Hello Simon,
> 
> On 13-10-14 07:14, Simon Glass wrote:
> >Hi Jeroen,
> >
> >On 12 October 2014 10:13, Jeroen Hofstee <jeroen@myspectrum.nl> wrote:
> >
> >>Hello Hans,
> >>
> >>On 12-10-14 12:25, Hans de Goede wrote:
> >>
> >>>Hi,
> >>>
> >>>This one seems to have fallen through the cracks.
> >>>
> >>>Regards,
> >>>
> >>>Hans
> >>>
> >>>  (for U-boot)
> >>nope, you replace an innocent warning (_might_ be) with
> >>bad code, without any comment it is just because gcc failed
> >>to recognize it is fine. Nor did you respond to the suggestion
> >>if it helps gcc to recognize that if the two booleans are merged
> >>into a single one. [or even split it in an if () if ()]. With this patch
> >>you prevent any serious warning in case the variable is actually
> >>used but not initialized, which is even worse if you ask me.
> >>
> >That is a pretty acerbic tone to take on the U-Boot list at least. Are you
> >two drinking buddies or something?
> 
> no, it is because we have discussed this patch before and resending
> it won't address the issue raised. But you are right, it is likely done with
> less evil intends then I took it for, so let me explain my concern again
> in a politer way. The problem is that gcc 4.9 starts warning in the
> following case:
> 
> int *ptr;
> 
> if (a)
>     ptr = something;
> 
> if (a && b)
>     ptr->bla = value;
> else
>    do_something_else();
> 
> 
> it will warn that ptr _might_ be used uninitialized (but it always is).
> This is fixed in this patch by assigning NULL to ptr, and while that makes
> the warning go away it actually prevents the valid warning, ptr _is_ used
> uninitialized if you start using it in the else case. Hence my request if we
> can't find a better solution for this.
> 
> Does anyone know a better solution for this or should we consider
> disabling the might be unused warning?

Frankly, looking at the code, this is a compiler bug since as you note
the pointer will always be initalized.  Since we share this code as-is
with upstream kernel, we should see if there's any interst there in
trying to re-write the code so that it's (roughly):
if (a)
 ptr = valid;

if (a && b && ptr)
 ptr->foo = bar;

Or if this gets the required "compiler is being stupid, file a bug"
volume required.

-- 
Tom

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

[U-Boot] [PATCH resend] kconfig: Fix compiler warning in menu.c

[Tom Rini]

On Mon, Oct 13, 2014 at 08:48:39AM +0200, Jeroen Hofstee wrote:
> Hello Simon,
> 
> On 13-10-14 07:14, Simon Glass wrote:
> >Hi Jeroen,
> >
> >On 12 October 2014 10:13, Jeroen Hofstee <jeroen@myspectrum.nl> wrote:
> >
> >>Hello Hans,
> >>
> >>On 12-10-14 12:25, Hans de Goede wrote:
> >>
> >>>Hi,
> >>>
> >>>This one seems to have fallen through the cracks.
> >>>
> >>>Regards,
> >>>
> >>>Hans
> >>>
> >>>  (for U-boot)
> >>nope, you replace an innocent warning (_might_ be) with
> >>bad code, without any comment it is just because gcc failed
> >>to recognize it is fine. Nor did you respond to the suggestion
> >>if it helps gcc to recognize that if the two booleans are merged
> >>into a single one. [or even split it in an if () if ()]. With this patch
> >>you prevent any serious warning in case the variable is actually
> >>used but not initialized, which is even worse if you ask me.
> >>
> >That is a pretty acerbic tone to take on the U-Boot list at least. Are you
> >two drinking buddies or something?
> 
> no, it is because we have discussed this patch before and resending
> it won't address the issue raised. But you are right, it is likely done with
> less evil intends then I took it for, so let me explain my concern again
> in a politer way. The problem is that gcc 4.9 starts warning in the
> following case:
> 
> int *ptr;
> 
> if (a)
>     ptr = something;
> 
> if (a && b)
>     ptr->bla = value;
> else
>    do_something_else();
> 
> 
> it will warn that ptr _might_ be used uninitialized (but it always is).
> This is fixed in this patch by assigning NULL to ptr, and while that makes
> the warning go away it actually prevents the valid warning, ptr _is_ used
> uninitialized if you start using it in the else case. Hence my request if we
> can't find a better solution for this.
> 
> Does anyone know a better solution for this or should we consider
> disabling the might be unused warning?

Frankly, looking at the code, this is a compiler bug since as you note
the pointer will always be initalized.  Since we share this code as-is
with upstream kernel, we should see if there's any interst there in
trying to re-write the code so that it's (roughly):
if (a)
 ptr = valid;

if (a && b && ptr)
 ptr->foo = bar;

Or if this gets the required "compiler is being stupid, file a bug"
volume required.

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20141013/7922ed89/attachment.pgp>