$NetBSD: README.warnings,v 1.5 2023/07/14 19:46:25 mrg Exp $
What to do about GCC warnings and NetBSD.
New GCC releases always come with a host of new warnings and
each new warning can find real bugs, find odd code, or simply
be a pain of new useless warnings, or all three and more.
As each warning has its own set of issues they each have their
own section and it is expected that this document will be
modified for updates to warnings and new warnings.
<bsd.own.mk> provides several variables for use in Makefiles:
COPTS.foo.c += ${CC_WNO_FORMAT_TRUNCATION}
COPTS.foo.c += ${CC_WNO_FORMAT_OVERFLOW}
COPTS.foo.c += ${CC_WNO_STRINGOP_OVERFLOW}
COPTS.foo.c += ${CC_WNO_STRINGOP_TRUNCATION}
COPTS.foo.c += ${CC_WNO_CAST_FUNCTION_TYPE}
COPTS.foo.c += ${CC_WNO_IMPLICIT_FALLTHROUGH}
COPTS.foo.c += ${CC_WNO_ADDRESS_OF_PACKED_MEMBER}
COPTS.foo.c += ${CC_WNO_MAYBE_UNINITIALIZED}
COPTS.foo.c += ${CC_WNO_RETURN_LOCAL_ADDR}
new GCC 10 warnings:
GCC 10 switched the default from "-fcommon" to "-fno-common",
which can cause multiply defined symbol issues. Ideally we
fix all of these, but "-fcommon" can be used otherwise.
-Wno-maybe-uninitialized
This warning was introduced in an ancient GCC but was
significantly enhanced in GCC 10, unfortunately, many of
the new instances are incorrect.
bsd.own.mk variable: ${CC_WNO_MAYBE_UNINITIALIZED}
-Wno-return-local-addr
This warning was introduced in GCC 5 and was enhanced in GCC
10. Unfortunately, the new instances are failing to correctly
analyze code flow and miss that most of the are handled.
bsd.own.mk variable: ${CC_WNO_RETURN_LOCAL_ADDR}
new GCC 9 warnings:
-Wno-address-of-packed-member
This warning was introduced in GCC 8.
This warning is similar to -Wformat-truncation, but for the
general family of string functions (str*(), etc.), and has
similar issues of false positives.
bsd.own.mk variable: ${CC_WNO_ADDRESS_OF_PACKED_MEMBER}
new GCC 8 warnings:
-Wstringop-truncation
This warning was introduced in GCC 8.
This warning is similar to -Wformat-truncation, but for the
general family of string functions (str*(), etc.), and has
similar issues of false positives.
bsd.own.mk variable: ${CC_WNO_STRINGOP_TRUNCATION}
-Wcast-function-type
This warning was introduced in GCC 8.
This warning can find real problems. Most instances are
false positives, and hopefully this warning will become
more useful in the future. See __FPTRCAST().
bsd.own.mk variable: ${CC_WNO_CAST_FUNCTION_TYPE}
new GCC 7 warnings:
-Wstringop-overflow
This warning was introduced in GCC 7.
This warning can find issues where source length is
passed as destination length (eg, strncpy() where the
length is strlen(src)) that are potential buffer overflow
cases and should always be inspected, but false positives
are also seen.
bsd.own.mk variable: ${CC_WNO_STRINGOP_OVERFLOW}
-Wformat-truncation
This warning was introduced in GCC 7.
This warning has many false positives where truncation is
either expected or unavoidable, but also finds several real
code bugs.
Code should always be manually inspected for this warning
as it does pick up real issues.
bsd.own.mk variable: ${CC_WNO_FORMAT_TRUNCATION}
-Wformat-overflow
This warning was introduced in GCC 7.
This warning typically identifies a real problem, but it may
fail to notice the code handles this case.
Code should always be manually inspected for this warning
as it does pick up real issues.
bsd.own.mk variable: ${CC_WNO_FORMAT_OVERFLOW}
-Wimplicit-fallthrough
This warning was introduced in GCC 7.
This warning has many false positives in GCC 7, and many are
fixed in newer GCC 10. Old uses should be checked occasionally.
Code should always be manually inspected for this warning
as it does pick up real issues.
bsd.own.mk variable: ${CC_WNO_IMPLICIT_FALLTHROUGH}
721b426c77