]> git.ipfire.org Git - thirdparty/qemu.git/commitdiff
compiler.h: Don't use compile-time assert when __NO_INLINE__ is defined
authorZenghui Yu <yuzenghui@huawei.com>
Wed, 5 Feb 2020 14:15:45 +0000 (22:15 +0800)
committerRichard Henderson <richard.henderson@linaro.org>
Fri, 28 Feb 2020 18:58:41 +0000 (10:58 -0800)
Our robot reported the following compile-time warning while compiling
Qemu with -fno-inline cflags:

In function 'load_memop',
    inlined from 'load_helper' at /qemu/accel/tcg/cputlb.c:1578:20,
    inlined from 'full_ldub_mmu' at /qemu/accel/tcg/cputlb.c:1624:12:
/qemu/accel/tcg/cputlb.c:1502:9: error: call to 'qemu_build_not_reached' declared with attribute error: code path is reachable
         qemu_build_not_reached();
         ^~~~~~~~~~~~~~~~~~~~~~~~
    [...]

It looks like a false-positive because only (MO_UB ^ MO_BSWAP) will
hit the default case in load_memop() while need_swap (size > 1) has
already ensured that MO_UB is not involved.

So the thing is that compilers get confused by the -fno-inline and
just can't accurately evaluate memop_size(op) at compile time, and
then the qemu_build_not_reached() is wrongly triggered by (MO_UB ^
MO_BSWAP).  Let's carefully don't use the compile-time assert when
no functions will be inlined into their callers.

Reported-by: Euler Robot <euler.robot@huawei.com>
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Zenghui Yu <yuzenghui@huawei.com>
Message-Id: <20200205141545.180-1-yuzenghui@huawei.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
include/qemu/compiler.h

index 85c02c16d3ad04d2790a437e1ff23795f84dac0d..c76281f3540bf07a13c5608c68256e72c2e54aec 100644 (file)
  * supports QEMU_ERROR, this will be reported at compile time; otherwise
  * this will be reported at link time due to the missing symbol.
  */
-#ifdef __OPTIMIZE__
+#if defined(__OPTIMIZE__) && !defined(__NO_INLINE__)
 extern void QEMU_NORETURN QEMU_ERROR("code path is reachable")
     qemu_build_not_reached(void);
 #else