Centos 7. SPH_UDF_ARGS. UDF arguments not right initialized

[ousharovsky]

Describe the bug
SPH_UDF_ARGS arguments not right initialized in manticore v 6.0.4 for Centos 7 and 10 million documents index. In Centos 8 and Ubuntu 20.04 all works fine.
There is also no problem for an index of 5 documents for Centos 7.

SPH_UDF_ARGS::arg_values has random or empty values for UDF threads.
SPH_UDF_ARGS::str_lengths may by negative value.

Describe the environment:
searchd --version
Manticore 6.0.4 4c99009@230601
Copyright (c) 2001-2016, Andrew Aksyonoff
Copyright (c) 2008-2016, Sphinx Technologies Inc (http://sphinxsearch.com)
Copyright (c) 2017-2023, Manticore Software LTD (https://manticoresearch.com)

uname -a
Linux host.domain 3.10.0-1160.88.1.el7.x86_64 #1 SMP Tue Mar 7 15:41:52 UTC 2023 x86_64 x86_64 x86_64 GNU/Linux

[sanikolaev]

@ousharovsky Thanks for filing the issue. Is it possible to get a sample UDF which demonstrates the problem?

[ousharovsky]

#include "sphinxudf.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#ifdef _MSC_VER
#define snprintf _snprintf
#define DLLEXPORT __declspec(dllexport)
#else
#define DLLEXPORT
#endif

/// UDF logging callback
/// write any message into daemon's log.
sphinx_log_fn* sphlog = NULL;

#ifndef NDEBUG
DLLEXPORT void libexactsearch_plugind_setlogcb(sphinx_log_fn* cblog) {
#else
DLLEXPORT void libexactsearch_plugin_setlogcb(sphinx_log_fn* cblog) {
#endif
  sphlog = cblog;
}

void UdfLog(char* szMsg) {
  if (sphlog)
    (*sphlog)(szMsg, -1);
}

#ifndef NDEBUG
DLLEXPORT int libexactsearch_plugind_ver() {
#else
DLLEXPORT int libexactsearch_plugin_ver() {
#endif
  return SPH_UDF_VERSION;
}

DLLEXPORT int strtoint_init(SPH_UDF_INIT* init, SPH_UDF_ARGS* args, char* error_message) {
  UdfLog("Called strtoint_init");
  if (args->arg_count != 1 || args->arg_types[0] != SPH_UDF_TYPE_STRING) {
    snprintf(error_message, SPH_UDF_ERROR_LEN, "STRTOINT() requires 1 string argument");
    return 1;
  }
  return 0;
}

DLLEXPORT sphinx_int64_t strtoint(SPH_UDF_INIT* init, SPH_UDF_ARGS* args, char* error_flag) {
  UdfLog("Called strtoint");
  const char* s = args->arg_values[0];
  int len = args->str_lengths[0], res = 0;

  // looks strange, but let's just take sum of digits, i.e. '123' -> 1+2+3 = 6.
  while (len > 0 && *s >= '0' && *s <= '9') {
    res += *s - '0';
    --len;
    ++s;
  }
  return res;
}

select strtoint('123') from big_index;
+-----------------+
| strtoint('123') |
+-----------------+
|               6 |
|               0 |
|               6 |
|               6 |
|               0 |
|               0 |
|               0 |
|               0 |
|               0 |
|               0 |
|               0 |
|               6 |
|               6 |
|               6 |
|               6 |
|               6 |
|               0 |
|               6 |
|               0 |
|               0 |
+-----------------+
20 rows in set (0,01 sec)

from CMakeList.txt:

set(MODULE_NAME exactsearch_plugin)

[sanikolaev]

Thanks. We'll look into this.