Page 257 - 《软件学报》2024年第6期

向毅 等: 基于多样性 SAT 求解器和新颖性搜索的软件产品线测试 2833

盖率的表现优于 SATNS( ↑ ), 且二者的差异幅度被评估为 large (l). 从表 3 可看出, 对覆盖率指标和 N=4, 6 和 10 而
言, 在至少 60% 的特征模型上 dSATNS 与 SATNS 具有不可忽视的差异. 此外, 无论差异幅度是 l, m 还是 s,
dSATNS 表现更优的模型所占百分比始终 (远) 高于其表现更差的模型所占百分比. 当 N=50 和 100 时, 在大多数
模型上 (60%+), 两算法的差异是可忽略的. 考虑缺陷检测率, dSATNS 表现更优的模型所占百分比始终高于
SATNS 表现更优的模型所占百分比. 由此可见, 在更多情形下, dSATNS 改进了 SATNS, 而且这种改进在相当比
例的特征模型上是本质性的 (即效应量幅度为 l 和 m).

100 100
• ◦ ± 88 92 • ◦ ± 86
80 84
80 80
72
68 66
覆盖率 (%) 60 40 56 44 54 缺陷检测率 (%) 60
40
28 40 28
24
20 20 18 14
12
6 6 6
4 2 4 4 4 4 2 2 2
0 0
N=4 N=6 N=10 N=50 N=100 N=4 N=6 N=10 N=50 N=100
(a) 覆盖率 (b) 缺陷检测率
图 2 根据 U 检验, dSATNS 显著优于 (•)、差于 (◦) 和等同于 (±)SATNS 的特征模型所占百分比

表 2 dSATNS 和 SATNS 两算法缺陷检测率的比较 (%)

N=4 N=6 N=10 N=50 N=100
FM
dSATNS SATNS dSATNS SATNS dSATNS SATNS dSATNS SATNS dSATNS SATNS
CounterStrikeSFM 66.33 66.50 ± 73.67 75.00 ◦ 83.33 84.00 ± 97.33 97.83 ± 99.33 99.33 ±
HiPAcc 55.50 55.75 ◦ 65.50 64.38 ± 75.00 74.75 ± 93.38 93.13 ± 96.63 96.75 ±
SPLSSimuelESPnP 64.75 64.25 ± 74.25 65.69 ± 99.75 ±
99.75
98.25 ±
73.25 ±
84.00
98.50
83.75 ±
JavaGC 54.75 55.75 ± 62.75 63.25 ± 71.75 71.13 ± 90.63 90.50 ± 95.00 94.88 ±
Polly 57.70 57.60 ± 65.80 65.60 ± 75.50 75.30 ± 93.50 93.20 ± 96.80 96.60 ±
DSSample 47.90 48.40 ± 55.20 55.00 ± 63.70 63.50 ± 86.60 86.60 ± 91.80 91.80 ±
VP9 54.50 54.70 ± 63.20 62.80 ± 71.90 71.40 ± 91.60 91.40 ± 95.80 96.00 ±
WebPortal 58.50 58.40 ± 67.10 66.70 ± 76.80 76.40 ± 94.60 94.20 ± 97.20 97.30 ±
JHipster 62.50 62.70 ± 70.60 70.40 ± 79.40 79.70 ± 96.00 95.80 ± 98.20 98.00 ±
Drupal 56.80 56.20 ± 65.00 64.40 ± 74.30 74.30 ± 94.50 94.00 • 97.90 97.80 ±
SmartHomev2.2 54.21 53.64 • 62.57 61.86 • 72.36 71.79 ± 93.14 92.79 ± 97.21 97.29 ±
VideoPlayer 62.63 62.13 ± 70.13 70.19 ± 79.50 79.13 ± 96.63 96.44 ± 98.94 98.88 ±
Amazon 47.25 47.38 ± 52.94 53.31 ± 60.19 60.25 ± 80.94 80.63 ± 86.81 86.94 ±
ModelTransformation 53.17 53.00 ± 61.94 61.50 ± 71.33 71.22 ± 92.33 92.44 ± 96.44 96.44 ±
CocheEcologico 65.45 65.45 ± 71.95 72.15 ± 79.20 79.35 ± 93.60 93.70 ± 96.35 96.30 ±
Printers 59.89 59.03 • 66.03 73.31 73.22 ± 89.89 89.83 ± 93.81 93.89 ±
fiasco_17_10 54.29 53.85 • 59.38 59.35 ± 66.81 66.67 ± 80.08 80.23 ± 83.50 83.81 ◦
uClibc-ng_1_0_29 44.28 43.83 • 50.07 49.52 • 56.80 56.30 ± 74.78 74.74 ± 80.70 81.04 ±
E-shop 50.45 49.53 • 58.50 57.30 • 67.75 67.82 ± 90.68 90.23 • 95.58 95.57 ±
toybox 90.10 89.52 • 93.96 93.64 • 96.76 96.79 ± 99.82 99.82 ± 99.96 99.96 ±
axTLS 83.24 83.14 ± 88.67 88.49 ± 93.38 93.26 ± 99.23 99.20 ± 99.77 99.76 ±
financial 46.21 46.41 ◦ 49.14 49.23 ± 53.51 53.43 ± 65.72 65.69 ± 71.19 71.17 ±
busybox_1_28_0 43.90 42.48 • 51.45 50.65 • 61.16 60.42 • 86.90 86.57 • 93.57 93.33 •
mpc50 51.12 51.14 ± 58.28 58.16 ± 66.89 67.54 ◦ 88.12 88.03 ± 93.13 92.97 ±
ref4955 50.69 50.73 ± 57.95 57.82 ± 67.01 67.04 ± 87.61 87.61 ± 92.88 92.83 ±
Linux 51.40 51.51 ± 58.52 58.31 ± 66.82 67.17 ± 87.43 87.46 ± 92.52 92.51 ±