https://bugs.dpdk.org/show_bug.cgi?id=1986
Bug ID: 1986
Summary: dpaax: build warnings with clang-23
Product: DPDK
Version: 26.11
Hardware: All
OS: All
Status: UNCONFIRMED
Severity: minor
Priority: Normal
Component: ethdev
Assignee: [email protected]
Reporter: [email protected]
Target Milestone: ---
Clang 23 checks for set but not used global variables.
../drivers/common/dpaax/dpaa_of.c:14:12: error: variable 'alive' set but not
used
[-Werror,-Wunused-but-set-global]
14 | static int alive;
| ^
clang-23 extends -Wunused-but-set-variable to file-scope statics
(-Wunused-but-set-global), which is enabled by -Wall and fatal under -Werror.
Earlier clang and GCC releases do not diagnose this case, so the code builds
there.
alive is assigned in of_init_path() and of_finish(), and is otherwise
referenced
only as the condition argument to DPAAX_HWWARN(), at 13 call sites in
dpaa_of.c.
That macro is defined in drivers/common/dpaax/dpaax_logs.h:
#ifdef RTE_LIBRTE_DPAAX_DEBUG
#define DPAAX_HWWARN(cond, fmt, ...) \
do {\
if (cond) \
DPAAX_LOG(DEBUG, "WARN: " fmt, ##__VA_ARGS__); \
} while (0)
#else
#define DPAAX_HWWARN(cond, fmt, ...) do { } while (0)
#endif
The disabled variant discards cond entirely, so alive is never read — hence
the
warning.
There are three options:
1. Remove all this code, since RTE_LIBRTE_DPAAX_DEBUG is not defined anywhere
and there is no meson build to set it. I.e dead code.
2. Add 0 to the do { } while() like:
#ifdef RTE_LIBRTE_DPAAX_DEBUG
#define DPAAX_HWWARN(cond, fmt, ...) \
do {\
if (cond) \
DPAAX_LOG(DEBUG, "WARN: " fmt, ##__VA_ARGS__); \
} while (0)
#else
#define DPAAX_HWWARN(cond, fmt, ...) \
do {\
if (0 && (cond)) \
break; \
} while (0)
#endif
3. Cleanup and keep the check.
Rename the variable because 'alive' is not descriptive.
Always do the check (no #ifdef) since it is good bug trap.
Make it a warning or notice level since buggy device tree is worth checking.
That would end up renaming or removing macro.
--
You are receiving this mail because:
You are the assignee for the bug.
