annoying new gcc 4.6.0 warnings.

There might be some valid bugs found (I think DaveM found a few already) from this new warning, but it seems like everything I've looked at so far is just noise.

Any thoughts ?

Quoted 5 lines · Dave Jones

There might be some valid bugs found (I think DaveM found a few already) from this new warning, but it seems like everything I've looked at so far is just noise.

Any thoughts ?

Yeah, let's just disable the crazy warning. Maybe some of them are valid, but it's like the f*cking sign-compare warning: most of them are just inane noise, and as such the warning is not worth the pain.

I assume (hope) that there is some -Wno-unused-but-set-variable thing we could do, the same way we do -Wno-trigraphs for another totally useless gcc warning.

Would it make any sense to add a config option to enable the spew for those that wanted to see it?

Quoted 2 lines · David Daney

Would it make any sense to add a config option to enable the spew for those that wanted to see it?

Add it to make W=1 instead?
See commit 4a5838ad9d2d

Quoted 6 lines · Joe Perches

Quoted 2 lines · David Daney

Would it make any sense to add a config option to enable the spew for those that wanted to see it?

Add it to make W=1 instead?
See commit 4a5838ad9d2d

This patch does that. Though I question the usefulness of W=1. I just built powernow-k8.o with W=1, and got 4030 lines of output. That's just insane.

My gcc (4.3.2) barfed over -Wno-unused-but-set-variable. So your patch needs to check if the option is supported.

Quoted 2 lines · Sam Ravnborg

My gcc (4.3.2) barfed over -Wno-unused-but-set-variable. So your patch needs to check if the option is supported.

I don't have a gcc older than 4.4.0. Can you try this ?

Dave
--

Disable the new -Wunused-but-set-variable that was added in gcc 4.6.0 It produces more false positives than useful warnings.

This can still be enabled using W=1

Signed-off-by Dave Jones<[email protected]>

Makefile +2 -0

index b967b96..29e16f2 100644
@@ -559,6 +559,8 @@ ifndef CONFIG_CC_STACKPROTECTOR
  KBUILD_CFLAGS += $(call cc-option, -fno-stack-protector)
  endif
 
+KBUILD_CFLAGS += $(call cc-option, -Wno-unused-but-set-variable)
+
  ifdef CONFIG_FRAME_POINTER
  KBUILD_CFLAGS	+= -fno-omit-frame-pointer -fno-optimize-sibling-calls
  else

scripts/Makefile.build +0 -0

index d5f925a..30627ab 100644

Quoted 1 line

+KBUILD_EXTRA_WARNINGS += -Wunused-but-set-variable

Well I didn't test it, but presumably if you need the $(call cc-option, -Wno-unused-but-set-variable), you would need similar here.

Quoted 4 lines · David Daney

Quoted 1 line

+KBUILD_EXTRA_WARNINGS += -Wunused-but-set-variable

Well I didn't test it, but presumably if you need the $(call cc-option, -Wno-unused-but-set-variable), you would need similar here.

Third time's the charm ?

Seems soo. It semi works now.
"make W=1" is broken for reasons that is not related to this patch - will submit a fix in a minute for that.

So you can add:

Acked-by Sam Ravnborg <[email protected]>

Tested-by Sam Ravnborg <[email protected]>

on the patch.

And if you really want to make me happy then you add a comment about _why_ we disable the warning.

Quoted 9 lines · Sam Ravnborg

Seems soo. It semi works now.
"make W=1" is broken for reasons that is not related to this patch - will submit a fix in a minute for that.

So you can add:

Acked-by Sam Ravnborg <[email protected]>

Tested-by Sam Ravnborg <[email protected]>

Quoted 4 lines

on the patch.

And if you really want to make me happy then you add a comment about _why_ we disable the warning.

ok, last time for reals.

Disable the new -Wunused-but-set-variable that was added in gcc 4.6.0 It produces more false positives than useful warnings.

This can still be enabled using W=1

Signed-off-by Dave Jones <[email protected]>

Acked-by Sam Ravnborg <[email protected]>

Tested-by Sam Ravnborg <[email protected]>

Makefile KBUILD_CFLAGS   := -Wall -Wundef -Wstrict-prototypes -Wno-trigraphs \ · line 363

Quoted 7 lines · Dave Jones

Makefile +2 -1

@@ -359,7 +359,8 @@ KBUILD_CFLAGS   := -Wall -Wundef -Wstrict-prototypes -Wno-trigraphs \
                   -fno-strict-aliasing -fno-common \
                   -Werror-implicit-function-declaration \
                   -Wno-format-security \
-                  -fno-delete-null-pointer-checks
+                  -fno-delete-null-pointer-checks \
+                  -Wno-unused-but-set-variable

Does this work ok for older gcc's? Do they understand that -Wno-unused-by-set-variable?

gcc-4.5.1 seems to understand it, but what about much older gccs?

gcc 4.6.0 enables a new warning by default, which spews ~3000 lines of extra warnings to my build. The warning looks like ..

arch/x86/kernel/cpu/cpufreq/powernow-k8.c: In function ‘pending_bit_stuck’:
arch/x86/kernel/cpu/cpufreq/powernow-k8.c:108:10: warning: variable ‘hi’ set but not used [-Wunused-but-set-variable] arch/x86/kernel/cpu/cpufreq/powernow-k8.c: In function ‘core_voltage_pre_transition’:
arch/x86/kernel/cpu/cpufreq/powernow-k8.c:340:14: warning: variable ‘lo’ set but not used [-Wunused-but-set-variable]

In those cases, there's no bug there, the code just doesn't use all the arguments
that get passed to rdmsr... For eg,

static int pending_bit_stuck(void)

{
        u32 lo, hi;

        if (cpu_family == CPU_HW_PSTATE)
                return 0;

        rdmsr(MSR_FIDVID_STATUS, lo, hi); 
        return lo & MSR_S_LO_CHANGE_PENDING ? 1 : 0; 
}

gcc's manpage says to dismiss this warning with attribute(unused), but sprinkling thousands of those through the source seems pretty ugly. As does rewriting functions like rdmsr to handle NULL as an argument.

There might be some valid bugs found (I think DaveM found a few already) from this new warning, but it seems like everything I've looked at so far is just noise.

Any thoughts ?

Dave

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

This new gcc-4.6.0 behavior can also cause build failures for archs which use -Werror by default (sparc, ppc) [1][2].

[1] https://lkml.org/lkml/2011/4/19/383
[2] https://lkml.org/lkml/2011/4/18/287