[llvm-bugs] [Bug 38789] New: warnings for untaken branch of ?: expression, at file scope
via llvm-bugs
llvm-bugs at lists.llvm.org
Fri Aug 31 05:19:45 PDT 2018
https://bugs.llvm.org/show_bug.cgi?id=38789
Bug ID: 38789
Summary: warnings for untaken branch of ?: expression, at file
scope
Product: clang
Version: unspecified
Hardware: PC
OS: Linux
Status: NEW
Severity: enhancement
Priority: P
Component: -New Bugs
Assignee: unassignedclangbugs at nondot.org
Reporter: arnd at linaro.org
CC: llvm-bugs at lists.llvm.org
I have encountered a couple of warnings in the Linux kernel for range checking
of shift counts. The common macro
#define DMA_BIT_MASK(n) (((n) == 64) ? ~0ULL : ((1ULL<<(n))-1))
produces a warning at file scope for a '64' argument, but not inside of
a function. I tried using __builtin_choose_expr() to avoid that, which
made it worse: that version always warns, regardless of the scope (it
also does this with gcc).
The same thing happened with a less common macro, which I've simplifed
into 'COND_BIT()' in the test case below.
The behavior is the same for all clang versions I tried (3.8 through 8),
but my expectation was to see no warning at all.
$ clang-8 -Wall -O2 -c -xc - << EOF
#define DMA_BIT_MASK(n) (((n) == 64) ? ~0ULL : ((1ULL<<(n))-1))
#define DMA_BIT_MASK_2(n) __builtin_choose_expr((n) == 64, ~0ULL,
(1ULL<<(n))-1)
#define COND_BIT(x) (((x) >= 0) ? (1ul << x) : 0ul)
#define COND_BIT_2(x) __builtin_choose_expr(x >= 0, 1ul << x, 0ul)
// warning: shift count >= width of type [-Wshift-count-overflow]
long long a = DMA_BIT_MASK(64);
// warning: shift count >= width of type [-Wshift-count-overflow]
long long b = DMA_BIT_MASK_2(64);
// warning: shift count is negative [-Wshift-count-negative]
// [gcc: no warning]
unsigned long c = COND_BIT(-1);
// warning: shift count is negative [-Wshift-count-negative]
unsigned long d = COND_BIT_2(-1);
int f(void)
{
// no warning
long long a = DMA_BIT_MASK(64);
// warning: shift count >= width of type [-Wshift-count-overflow]
long long b = DMA_BIT_MASK_2(64);
// no warning
unsigned long c = COND_BIT(-1);
// warning: shift count is negative [-Wshift-count-negative]
unsigned long d = COND_BIT_2(-1);
// prevent unused variable wanings
return a & b & c & d;
}
EOF
--
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20180831/fb359de6/attachment.html>
More information about the llvm-bugs
mailing list