I ran into the following build failure with gcc 4.6:
bench/sched-pipe.c: In function 'bench_sched_pipe':
bench/sched-pipe.c:58:6: error: variable 'ret' set but not used
[-Werror=unused-but-set-variable]
cc1: all warnings being treated as errors
As the main kernel Makefile has added -Wno-unused-but-set-variable
and sched-pipe.c claims:
/*
* why does "ret" exist?
* discarding returned value of read(), write()
* causes error in building environment for perf
*/
This patch creates a new perf/Makefile variable, LESS_WARNINGS and
adds -Wno-unused-but-set-variable. The new variable helps distinguish
between the added warnings in EXTRA_WARNINGS versus those we want to
remove, including the existing -Wno-system-headers which I moved to
LESS_WARNINGS.
Signed-off-by: Darren Hart <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Bruce Ashfield <[email protected]>
---
tools/perf/Makefile | 10 ++++++++--
1 files changed, 8 insertions(+), 2 deletions(-)
diff --git a/tools/perf/Makefile b/tools/perf/Makefile
index 032ba63..fdb35dc 100644
--- a/tools/perf/Makefile
+++ b/tools/perf/Makefile
@@ -74,7 +74,6 @@ EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wredundant-decls
EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wstrict-aliasing=3
EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wswitch-default
EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wswitch-enum
-EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wno-system-headers
EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wundef
EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wwrite-strings
EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wbad-function-cast
@@ -85,6 +84,13 @@ EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wold-style-definition
EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wstrict-prototypes
EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wdeclaration-after-statement
+#
+# Disable warnings that cause build problems or needlessly polute output:
+#
+
+LESS_WARNINGS := -Wno-system-headers
+LESS_WARNINGS := $(LESS_WARNINGS) -Wno-unused-but-set-variable
+
ifeq ("$(origin DEBUG)", "command line")
PERF_DEBUG = $(DEBUG)
endif
@@ -92,7 +98,7 @@ ifndef PERF_DEBUG
CFLAGS_OPTIMIZE = -O6
endif
-CFLAGS = -fno-omit-frame-pointer -ggdb3 -Wall -Wextra -std=gnu99 -Werror $(CFLAGS_OPTIMIZE) -D_FORTIFY_SOURCE=2 $(EXTRA_WARNINGS) $(EXTRA_CFLAGS)
+CFLAGS = -fno-omit-frame-pointer -ggdb3 -Wall -Wextra -std=gnu99 -Werror $(CFLAGS_OPTIMIZE) -D_FORTIFY_SOURCE=2 $(EXTRA_WARNINGS) $(LESS_WARNINGS) $(EXTRA_CFLAGS)
EXTLIBS = -lpthread -lrt -lelf -lm
ALL_CFLAGS = $(CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64
ALL_LDFLAGS = $(LDFLAGS)
--
1.7.1
Nevermind, this appears to be user error as the __used directive now
upstream should address this. Apologies.
--
Darren
On 06/02/2011 01:47 PM, Darren Hart wrote:
> I ran into the following build failure with gcc 4.6:
>
> bench/sched-pipe.c: In function 'bench_sched_pipe':
> bench/sched-pipe.c:58:6: error: variable 'ret' set but not used
> [-Werror=unused-but-set-variable]
> cc1: all warnings being treated as errors
>
> As the main kernel Makefile has added -Wno-unused-but-set-variable
> and sched-pipe.c claims:
>
> /*
> * why does "ret" exist?
> * discarding returned value of read(), write()
> * causes error in building environment for perf
> */
>
> This patch creates a new perf/Makefile variable, LESS_WARNINGS and
> adds -Wno-unused-but-set-variable. The new variable helps distinguish
> between the added warnings in EXTRA_WARNINGS versus those we want to
> remove, including the existing -Wno-system-headers which I moved to
> LESS_WARNINGS.
>
> Signed-off-by: Darren Hart <[email protected]>
> Cc: Ingo Molnar <[email protected]>
> Cc: Bruce Ashfield <[email protected]>
> ---
> tools/perf/Makefile | 10 ++++++++--
> 1 files changed, 8 insertions(+), 2 deletions(-)
>
> diff --git a/tools/perf/Makefile b/tools/perf/Makefile
> index 032ba63..fdb35dc 100644
> --- a/tools/perf/Makefile
> +++ b/tools/perf/Makefile
> @@ -74,7 +74,6 @@ EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wredundant-decls
> EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wstrict-aliasing=3
> EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wswitch-default
> EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wswitch-enum
> -EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wno-system-headers
> EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wundef
> EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wwrite-strings
> EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wbad-function-cast
> @@ -85,6 +84,13 @@ EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wold-style-definition
> EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wstrict-prototypes
> EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wdeclaration-after-statement
>
> +#
> +# Disable warnings that cause build problems or needlessly polute output:
> +#
> +
> +LESS_WARNINGS := -Wno-system-headers
> +LESS_WARNINGS := $(LESS_WARNINGS) -Wno-unused-but-set-variable
> +
> ifeq ("$(origin DEBUG)", "command line")
> PERF_DEBUG = $(DEBUG)
> endif
> @@ -92,7 +98,7 @@ ifndef PERF_DEBUG
> CFLAGS_OPTIMIZE = -O6
> endif
>
> -CFLAGS = -fno-omit-frame-pointer -ggdb3 -Wall -Wextra -std=gnu99 -Werror $(CFLAGS_OPTIMIZE) -D_FORTIFY_SOURCE=2 $(EXTRA_WARNINGS) $(EXTRA_CFLAGS)
> +CFLAGS = -fno-omit-frame-pointer -ggdb3 -Wall -Wextra -std=gnu99 -Werror $(CFLAGS_OPTIMIZE) -D_FORTIFY_SOURCE=2 $(EXTRA_WARNINGS) $(LESS_WARNINGS) $(EXTRA_CFLAGS)
> EXTLIBS = -lpthread -lrt -lelf -lm
> ALL_CFLAGS = $(CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64
> ALL_LDFLAGS = $(LDFLAGS)
--
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel
* Darren Hart <[email protected]> wrote:
> This patch creates a new perf/Makefile variable, LESS_WARNINGS and
> adds -Wno-unused-but-set-variable. The new variable helps
> distinguish between the added warnings in EXTRA_WARNINGS versus
> those we want to remove, including the existing -Wno-system-headers
> which I moved to LESS_WARNINGS.
Btw., while you resolved the issue, it would nevertheless nice to
introduce a 'make WERROR=0' build method that removes -Werror.
Would you be interested in sending a patch for that?
There would be several advantages of such a patch:
- GCC often introduces new warnings with lots of false positives -
breaking -Werror builds. WERROR=0 would allow one to build perf
without much fuss - while people would still send patches to avoid
the fuss of having to type WERROR=0 :-)
- bisecting back to commits that produce a (mostly harmless) warning
on some compiler is more difficult. With WERROR=0 one could bisect
without worrying about harmless warnings.
Thanks,
Ingo