[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[csmith-bugs] partially overlapping lvalues: bug still occurs as lv1 = (e, lv2);



Hello,

this is a follow-up to:
http://www.flux.utah.edu/listarchives/csmith-dev/msg00308.html

The same bug still seems to happen in the following program:

/*
 * This is a RANDOMLY GENERATED PROGRAM.
 *
 * Generator: csmith 2.1.0
 * Git version: a5e5fd1
 * Options:   --max-expr-complexity 15 --max-pointer-depth 3 --max-funcs 4 --max-array-dim 2 --max-array-len-per-dim 3 --max-struct-fields 12 --max-union-fields 12 --no-volatiles --bitfields --no-argc --unions
 * Seed:      3984942495
 */

I get:

t.c:211:[value] warning: Partially overlapping lvalue assignment "*l_756=l_46.f0.f10;". Left address in bits:
                 {{ l_46 -> {0} }}. Right address in bits: {{ l_46 -> {288} }}. assert(separated or same)

I think that the intention of the words "read from another object" in 6.5.16.1:3
is probably to encompass the statement found in that program:

(*l_756) = (... , l_46.f0.f10);

As with previous recurring bugs, this one has been much rarer after
the initial fixes. In fact, it is the only new finding for
8 Nehalem cores working at 2.67GHz full-time for the elapsed week.

Pascal