Discrepancy between BMH or BM result and Turbo result
agroce opened this issue · 3 comments
Found this using the DeepState tool for fuzzing, took an hour of fuzzing via AFL.
Here's a C++ program that shows the problem, also available at https://github.com/agroce/deepstate-boyer-moore-horspool/blob/master/bmhturbofail.cpp
The two string sizes are minimal, as are the byte values used (lower any of the bytes any more, or shrink the string, and there is at least no similar failing test case, so the exact pattern is meaningful).
#include <cstring>
#include <stdio.h>
#include <assert.h>
#include "Horspool.cpp"
#include "BoyerMooreAndTurbo.cpp"
int main() {
size_t needle_length = 17;
size_t haystack_length = 256;
unsigned char needle[1024];
unsigned char haystack[1024];
needle[0] = 255;
needle[1] = 255;
needle[2] = 16;
needle[3] = 16;
needle[4] = 16;
needle[5] = 16;
needle[6] = 16;
needle[7] = 16;
needle[8] = 16;
needle[9] = 64;
needle[10] = 16;
needle[11] = 16;
needle[12] = 255;
needle[13] = 255;
needle[14] = 16;
needle[15] = 16;
needle[16] = 16;
needle[17] = 0;
haystack[0] = 1;
haystack[1] = 1;
haystack[2] = 1;
haystack[3] = 1;
haystack[4] = 1;
haystack[5] = 1;
haystack[6] = 1;
haystack[7] = 1;
haystack[8] = 1;
haystack[9] = 1;
haystack[10] = 1;
haystack[11] = 1;
haystack[12] = 255;
haystack[13] = 255;
haystack[14] = 16;
haystack[15] = 16;
haystack[16] = 16;
haystack[17] = 255;
haystack[18] = 255;
haystack[19] = 16;
haystack[20] = 16;
haystack[21] = 16;
haystack[22] = 16;
haystack[23] = 16;
haystack[24] = 16;
haystack[25] = 16;
haystack[26] = 64;
haystack[27] = 16;
haystack[28] = 16;
haystack[29] = 255;
haystack[30] = 255;
haystack[31] = 16;
haystack[32] = 16;
haystack[33] = 16;
haystack[34] = 1;
haystack[35] = 1;
haystack[36] = 1;
haystack[37] = 1;
haystack[38] = 1;
haystack[39] = 1;
haystack[40] = 1;
haystack[41] = 1;
haystack[42] = 1;
haystack[43] = 1;
haystack[44] = 1;
haystack[45] = 1;
haystack[46] = 1;
haystack[47] = 1;
haystack[48] = 1;
haystack[49] = 1;
haystack[50] = 1;
haystack[51] = 1;
haystack[52] = 1;
haystack[53] = 1;
haystack[54] = 1;
haystack[55] = 1;
haystack[56] = 1;
haystack[57] = 1;
haystack[58] = 1;
haystack[59] = 1;
haystack[60] = 1;
haystack[61] = 1;
haystack[62] = 1;
haystack[63] = 1;
haystack[64] = 1;
haystack[65] = 1;
haystack[66] = 1;
haystack[67] = 1;
haystack[68] = 1;
haystack[69] = 1;
haystack[70] = 1;
haystack[71] = 1;
haystack[72] = 1;
haystack[73] = 1;
haystack[74] = 1;
haystack[75] = 1;
haystack[76] = 1;
haystack[77] = 1;
haystack[78] = 1;
haystack[79] = 1;
haystack[80] = 1;
haystack[81] = 1;
haystack[82] = 1;
haystack[83] = 1;
haystack[84] = 1;
haystack[85] = 1;
haystack[86] = 1;
haystack[87] = 1;
haystack[88] = 1;
haystack[89] = 1;
haystack[90] = 1;
haystack[91] = 1;
haystack[92] = 1;
haystack[93] = 1;
haystack[94] = 1;
haystack[95] = 1;
haystack[96] = 1;
haystack[97] = 1;
haystack[98] = 1;
haystack[99] = 1;
haystack[100] = 1;
haystack[101] = 1;
haystack[102] = 1;
haystack[103] = 1;
haystack[104] = 1;
haystack[105] = 1;
haystack[106] = 1;
haystack[107] = 1;
haystack[108] = 1;
haystack[109] = 1;
haystack[110] = 1;
haystack[111] = 1;
haystack[112] = 1;
haystack[113] = 1;
haystack[114] = 1;
haystack[115] = 1;
haystack[116] = 1;
haystack[117] = 1;
haystack[118] = 1;
haystack[119] = 1;
haystack[120] = 1;
haystack[121] = 1;
haystack[122] = 1;
haystack[123] = 1;
haystack[124] = 1;
haystack[125] = 1;
haystack[126] = 1;
haystack[127] = 1;
haystack[128] = 1;
haystack[129] = 1;
haystack[130] = 1;
haystack[131] = 1;
haystack[132] = 1;
haystack[133] = 1;
haystack[134] = 1;
haystack[135] = 1;
haystack[136] = 1;
haystack[137] = 1;
haystack[138] = 1;
haystack[139] = 1;
haystack[140] = 1;
haystack[141] = 1;
haystack[142] = 1;
haystack[143] = 1;
haystack[144] = 1;
haystack[145] = 1;
haystack[146] = 1;
haystack[147] = 1;
haystack[148] = 1;
haystack[149] = 1;
haystack[150] = 1;
haystack[151] = 1;
haystack[152] = 1;
haystack[153] = 1;
haystack[154] = 1;
haystack[155] = 1;
haystack[156] = 1;
haystack[157] = 1;
haystack[158] = 1;
haystack[159] = 1;
haystack[160] = 1;
haystack[161] = 1;
haystack[162] = 1;
haystack[163] = 1;
haystack[164] = 1;
haystack[165] = 1;
haystack[166] = 1;
haystack[167] = 1;
haystack[168] = 1;
haystack[169] = 1;
haystack[170] = 1;
haystack[171] = 1;
haystack[172] = 1;
haystack[173] = 1;
haystack[174] = 1;
haystack[175] = 1;
haystack[176] = 1;
haystack[177] = 1;
haystack[178] = 1;
haystack[179] = 1;
haystack[180] = 1;
haystack[181] = 1;
haystack[182] = 1;
haystack[183] = 1;
haystack[184] = 1;
haystack[185] = 1;
haystack[186] = 1;
haystack[187] = 1;
haystack[188] = 1;
haystack[189] = 1;
haystack[190] = 1;
haystack[191] = 1;
haystack[192] = 1;
haystack[193] = 1;
haystack[194] = 1;
haystack[195] = 1;
haystack[196] = 1;
haystack[197] = 1;
haystack[198] = 1;
haystack[199] = 1;
haystack[200] = 1;
haystack[201] = 1;
haystack[202] = 1;
haystack[203] = 1;
haystack[204] = 1;
haystack[205] = 1;
haystack[206] = 1;
haystack[207] = 1;
haystack[208] = 1;
haystack[209] = 1;
haystack[210] = 1;
haystack[211] = 1;
haystack[212] = 1;
haystack[213] = 1;
haystack[214] = 1;
haystack[215] = 1;
haystack[216] = 1;
haystack[217] = 1;
haystack[218] = 1;
haystack[219] = 1;
haystack[220] = 1;
haystack[221] = 1;
haystack[222] = 1;
haystack[223] = 1;
haystack[224] = 1;
haystack[225] = 1;
haystack[226] = 1;
haystack[227] = 1;
haystack[228] = 1;
haystack[229] = 1;
haystack[230] = 1;
haystack[231] = 1;
haystack[232] = 1;
haystack[233] = 1;
haystack[234] = 1;
haystack[235] = 1;
haystack[236] = 1;
haystack[237] = 1;
haystack[238] = 1;
haystack[239] = 1;
haystack[240] = 1;
haystack[241] = 1;
haystack[242] = 1;
haystack[243] = 1;
haystack[244] = 1;
haystack[245] = 1;
haystack[246] = 1;
haystack[247] = 1;
haystack[248] = 1;
haystack[249] = 1;
haystack[250] = 1;
haystack[251] = 1;
haystack[252] = 1;
haystack[253] = 1;
haystack[254] = 1;
haystack[255] = 1;
haystack[256] = 0;
const occtable_type occ = CreateOccTable(needle, needle_length);
const skiptable_type skip = CreateSkipTable(needle, needle_length);
size_t BMH_result = (int) SearchInHorspool(haystack, haystack_length, occ, needle, needle_length);
size_t Turbo_result = (int) SearchInTurbo(haystack, haystack_length, occ, skip, needle, needle_length);
printf("BMH: %zu", BMH_result);
printf("Turbo: %zu", Turbo_result);
assert(BMH_result == Turbo_result);
}
(I think the hour is optimistic, it may have gotten lucky; I'd fuzzed this code a while back and found nothing with a 24 hour run, and another run in progress hasn't found it yet)
@FooBarWidget let me know when you get a chance if this is real or a subtle issue with the API I'm somehow missing. It's genuinely a fairly hard target even for a fuzzer, without some patience. I know this project hasn't been touched in a while, but it's still pegged on your github and seems to have users!
Hi, sorry I'm not maintaining this repository at the moment. But if you're able to fix this bug and also add a test for it and submit a PR then I would be grateful.