Hello

Some time ago, I was at a presentation where Mark Scranton had a
script to capture the worst running SQL on a machine. Since he is busy
touring, I thought I would ask the community if anyone had that
particular script.

Thanks in advance for looking.


tom

Tom,

I don't know what script Mark Scranton had. But I have been using a
script a lot which you can find on my website:

http://michael-mueller-it.de/sqlstat.html

Michael

tomcaml@yahoo.com wrote:
> Hello
>
> Some time ago, I was at a presentation where Mark Scranton had a
> script to capture the worst running SQL on a machine. Since he is busy
> touring, I thought I would ask the community if anyone had that
> particular script.
>
> Thanks in advance for looking.
>
>
> tom
>

Thank you Michael. I will give it a whirl. Any gotchyas, environment
specifics, etc?

Tom

michael

i get an error reading header on this file after download


tom

Can you be a bit more specific? What error and on what operating system?

Michael


tomcaml@yahoo.com wrote:
> michael
>
> i get an error reading header on this file after download
>
>
> tom
>

Tom,

I'm attaching the major files. But I'm still interested to find out what
your downloading problem was.

Michael


tomcaml@yahoo.com wrote:
> michael
>
> i get an error reading header on this file after download
>
>
> tom
>


#!/usr/bin/ksh
if [ ! -x /usr/bin/ksh ]
then
echo "please change '#!/usr/bin/ksh' in line 1 to the path of a ksh"
echo "compatible shell and delete this if ..."
exit 1
fi

# Multiuser OnLine sql statistics using onstat commands (no sysmaster).
#
# Usage: SqlStat [ nsamples [ sleep-secs-per-sample [ nrepetitions ] ] ]
# or: SqlStat -h
#
# Results are in directory "erg.<hh:mm:ss>". There will be one such directory
# created for each repetition (parameter nrepetitions).
#
# Filename: erg.<hh:mm:ss>/0...0<num>.*_*
# <num> is the number of samples for the sql statement in this file.
# Just use "ls -l" to show files ordered by fequency of sql statements.
#
# If you have statements with varying constatns (no prepare, like
# "select ... where a = 123", "select ... where a = 987") use the
#
# use_scansql=y
#
# setting (and compile scansql.c), else you may want to use
# "use_scansql=n". The scansql program must be in the PATH if used
# (for example in the current directory if that's in the path).
#
# On some systems with huge shared memory (Solaris 8) starting and
# exiting onstat commands becomes very slow. Setting
#
# use_onstat_coprocess=y
#
# Makes the script communicate to a single onstat -i ksh coprocess.
# This option needs the cfilter.c program to be compiled and put in the PATH.

###### user settable options #############################################
#

# defaults for command line arguments nsamples, sleep-secs-per-sample
# and nrepetitions
nsamples=120
sleep=1
nrepetitions=1

# use scansql program = y/n
#use_scansql=y
use_scansql=n

# use onstat -i coprocess instead of many onstat commands, = y/n
#use_onstat_coprocess=y
use_onstat_coprocess=n

# output sql hostvariables (version >= 9.4), = y/n
#output_hv=y
output_hv=n

#
###### end of user settable options ######################################

if [ "$1" = -h ]
then
echo \
"Usage: SqlStat [ nsamples [ sleep-secs-per-sample [ nrepetitions ] ] ]"
echo \
" or: SqlStat -h"
echo \
" results will be put into directory erg.<hh:mm:ss>"
exit 0
fi

# assume most commands called here are in /usr/bin
export PATH=/usr/bin:$PATH

[ $# -ge 1 ] && nsamples="$1"
[ $# -ge 2 ] && sleep="$2"
[ $# -ge 3 ] && nrepetitions="$3"

if [ "$output_hv" = y ]
then
id | egrep 'uid=[0-9]*\(informix\)' >/dev/null 2>&1
if [ $? -ne 0 ]
then
echo "must run as user informix to use output_hv option"
exit 1
fi
fi

# remove temp files on exit
trap 'rm -f ../.SqlStat.tmp*' 0

# Finds program $1 in PATH and prints absolute path on stdout. Aborts
# if not found.
find_prog() {
h=`type "$1" 2>/dev/null`
e=$?
prog=`echo $h | awk '{print $NF}'`
[[ "$prog" != /* ]] && prog=`pwd`/$prog
if [ $e -ne 0 -o ! -x "$prog" ]
then
echo "cannot find program $1 in PATH"
exit 1
fi
echo $prog
}

# set $scansql either to cat or to absolute scansql program path
if [ "$use_scansql" = y ]
then
scansql=`find_prog scansql`
echo using $scansql
else
scansql=cat
fi

offmsg="shared memory not initialized for INFORMIXSERVER '$INFORMIXSERVER'"

# start onstat -i as a ksh coprocess and define onstat as a shell function
if [ "$use_onstat_coprocess" = y ]
then
cfilter=`find_prog cfilter`
echo using $cfilter

onstat -i |&
exec 6<&p 7>&p # save fds
onstatpid=$!
echo started coprocess onstat -i, pid $onstatpid

# temp file for cfilter program
ctmp=/tmp/.SqlStat.tmp99
rm -f $ctmp
if [ $? -ne 0 ]
then
echo cannot remove temp file $ctmp
exit 1
fi

trap '[ $onstatpid != 0 ] && kill -9 $onstatpid 2>/dev/null
rm -f ../.SqlStat.tmp* $ctmp' 0

# Function onstat() is supposed to behave like the original onstat
# command. It does it's job by communicating with the onstat -i
# coprocess using the external cfilter program. The shell always
# calls the function if defined.
onstat() {
if [ $onstatpid = 0 ] # coprocess already killed
then
echo "$offmsg"
return 1
fi

# export CFILTER_TRACE_FILE=/tmp/cfilter.trace
$cfilter "$*" $ctmp $onstatpid 4<&6 5>&7
# kill this job if oninit is down
# onstat -i is killed by trap 15
if [ $? -ne 0 ]
then
echo "$offmsg" 1>&2
kill $$ >/dev/null 2>&1
return 1 # not reached
fi

return 0
}
else # use_onstat_coprocess=n
onstat=`find_prog onstat`

onstat() {
$onstat $*
# kill this job if oninit is down
# but some older versions of onstat don't set the exit status!
if [ $? -ne 0 ]
then
echo "$offmsg" 1>&2
kill $$ >/dev/null 2>&1
return 1 # not reached
fi

return 0
}
fi

onstat_u () {
onstat -u | awk '$2 == "active," {exit}
f == 1 {print $1, $3}
$1 == "address" {f = 1}'
}

typeset -Z5 k=0

echo nsamples $nsamples, sleep-secs-per-sample $sleep, nrepetitions $nrepetitions

m=1
while [ $m -le $nrepetitions ]
do
echo "**** loop $m, `date`"

erg=erg.`date +%H:%M:%S`
mkdir $erg >/dev/null 2>&1
cd $erg
#rm -f *.*_*
> INFO
if [ "$use_onstat_coprocess" = y ]
then
echo "onstat -i coprocess $onstatpid" >> INFO
fi
echo "start: `date`, pid $$" >> INFO

num=0
n=1
while [ $n -le $nsamples ]
do
echo "$n of $nsamples "

# "rstcb | sid" Liste aller user
onstat_u | sort >../.SqlStat.tmp0

# find all interesting threads: rstcb, state, ...
onstat -g ath | egrep -v "netnorm|sm_read" | \
awk '/^ *[0-9].*sqlexec/ \
{printf("%s %s=%s=%s=%s=%s=%s\n", $3, $5, $6, $7, $8, $9, $10);}' \
>../.SqlStat.tmp1

i=`cat ../.SqlStat.tmp1 | wc -l`
if [ $i -gt 0 ]
then
# Choose a random thread
((i = (RANDOM % i ) + 1))
# get sid
sed -n -e "$i p" ../.SqlStat.tmp1 > ../.SqlStat.tmp2
sid=`join -o 2.2 ../.SqlStat.tmp2 ../.SqlStat.tmp0`

if [ "$sid" ]
then
# Get sql statement
onstat -g sql $sid > ../.SqlStat.tmp4
awk '/^Current SQL statement/ { f = 1; next; }
f == 1 { print; }
/^$/ { f = 0; } ' ../.SqlStat.tmp4 | \
$scansql >../.SqlStat.tmp3

if [ -s ../.SqlStat.tmp3 ]
then
((num += 1))
# file name: <counter>.<sum output 1>_<sum output 2>
set -- `sum ../.SqlStat.tmp3`
f=`ls *.$1_$2 2>/dev/null`
if [ "$f" ]
then
# file exists, increment counter part in filename
k=${f%%.*}
((k += 1))
f2=$k.$1_$2
mv $f $f2
else
echo >> ../.SqlStat.tmp3
let k=1
f2=$k.$1_$2
mv ../.SqlStat.tmp3 $f2
fi

# append time, sid, host, etc.
t=`date +%d.%m.%H:%M`
set -- `onstat -g ses $sid | \
awk 'f == 1 { print $2, $4, $5; exit; }
$1 == "id" { f = 1; }'`
if [ "$1" ]
then
fe=`ps -fp$2 | awk 'NR == 2 {print $8}'`
echo "@ $t, ses $sid, user $1, pid $2, host $3, fe $fe"\
>> $f2
echo " `awk '{print $2}' ../.SqlStat.tmp2`" | \
sed -e "s/=/ /g" >> $f2
if [ "$output_hv" = y ]
then # append host variables
awk '/^Host variables :/ { f = 1; next; }
f == 1 { print; }
/^$/ { if (f) exit; } ' ../.SqlStat.tmp4 \
>> $f2
fi
else
echo "@ $t: session exited" >> $f2
fi
fi
fi
fi

[ $sleep -gt 0 ] && sleep $sleep
((n += 1))
done

echo
echo "end: `date`" >> INFO
echo $num non empty samples >> INFO

((m += 1))

# for rm ../tmpfiles on exit avoid last cd ..
[ $m -le $nrepetitions ] && cd ..
done

/*
* Helper program for StatSql script to communicate with onstat -i ksh
* coprocess. Fd 5 is opened by the script for outputting commands into
* onstat -i and fd 4 for reading input from onstat -i.
* cfilter writes the command found in argv[1] to onstat -i followed by
* "- | tee /dev/null" which makes onstat -i flush it's output. Then it
* reads the onstat command's output until the next 'onstat> ' prompt is
* found. Any remaining input text is written to the argv[2] temp file for
* consumption by the next call. (Not needed for onstat 9.4, text ends with
* prompt. Needed for 7.24.UC8.) Signals SIGINT and redirected to the
* onstat -i process (with pid in argv[3]). They make onstat to interrupt
* it's running output and send the next prompt.
*
* All command output is written to fd 1 (stdout).
*
* Usage: cfilter "onstat-command-line" tmp-file-name
*
* To trace the program set env variable CFILTER_TRACE_FILE to a file pathname.
*
* Exit status: 0 = ok, 1 = shared memory not initialized, -1 = error.
*/

#include <stdio.h>
#include <stdarg.h>
#include <errno.h>
#include <string.h>
#include <stdlib.h>
#include <signal.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>

#define L (8*1024)
char iBuf[L + 1];
char oBuf[L + 1];

/* to grep for a string in onstat output */
struct findStr {
char *str;
char *p;
};

FILE *tfp = NULL; /* trace file */
int tmpFd;

int onstatPid;

volatile int wErr = 0; /* changed in signal handler */

volatile struct findStr onstatPrompt; /* changed in signal handler */
volatile struct findStr onstatState;

int FoundStr(volatile struct findStr *fstr, int c);
ssize_t Read(int fd, char *buf, size_t l);
ssize_t Write(int fd, char *buf, size_t l);
void Error(char *str, ...);
void handler(int sig);

int
main(int argc, char **argv)
{
int ifd = 4, ofd = 5;
int l, n, k, lCnt, nl, e;
char *p;

/*
volatile int dbg = 1;
fprintf(stderr, "cfilter: debug me, pid %d\n", getpid());
while (dbg) {
sleep(1);
}
*/

if (argc != 4)
Error("usage: %s onstat-command tmp-file-name onstat-pid",
argv[0]);

if (strlen(argv[1]) > L - 50)
Error("onstat-command too long");

onstatPid = atoi(argv[3]);
/* redirect SIGINT, SIGPIPE, make write() return EPIPE error */
sigset(SIGINT, handler);
sigset(SIGPIPE, handler);

/* trace file */
p = getenv("CFILTER_TRACE_FILE");
if (p) {
tfp = fopen(p, "a");
if (!tfp)
Error("cannot open trace file %s", p);
setbuf(tfp, NULL);
}
if (tfp)
fprintf(tfp, "\n=== onstat %s, tmp %s, pid %d ===\n",
argv[1], argv[2], onstatPid);

/* open temp file */
do
tmpFd = open(argv[2], O_RDWR);
while (tmpFd < 0 && errno == EINTR);

if (tfp && tmpFd >= 0)
fprintf(tfp, "opened temp file\n");

/* output command to onstat -i */
sprintf(oBuf, "%s\n- | tee /dev/null\n", argv[1]);
l = strlen(oBuf);
if (Write(ofd, oBuf, l) != l) {
if (tfp)
fprintf(tfp, "write to onstat -i failed, errno %d\n",
errno);
return 1; /* shared memory not init ?? */
}

/* read the onstat command's output */
if (tfp)
fprintf(tfp, "reading ...\n");

/* first read and skip the header from the previous command's
* "- | tee /dev/null" output. It's terminated by a "\nonstat> " prompt.
*/
onstatPrompt.str = "\nonstat> ";
onstatPrompt.p = onstatPrompt.str;
nl = 1;
while ((l = Read(ifd, iBuf, L)) > 0) {
iBuf[l] = 0;
for (n = 0; n < l; ++n) {
if (tfp) {
if (nl) {
fprintf(tfp, "0: ");
nl = 0;
}
fputc(iBuf[n], tfp);
if (iBuf[n] == '\n')
nl = 1;
}
if (FoundStr(&onstatPrompt, iBuf[n])) {
if (tfp)
fprintf(tfp, "\n");
++n;
goto loop2;
}
}
}
if (l < 0)
Error("read in loop 1");
Error("first onstat prompt not found");

/* now read and the onstat -i command's output and write it to fd 1.
* Output is again terminated by a "\nonstat> " prompt. If oninit
* was shutdown (and maybe rebooted) while onstat -i was running
* the state in the header output is "-- Initialization --".
* Abort in this case.
*/
loop2:
onstatPrompt.p = onstatPrompt.str;
onstatState.str = "-- Initialization --";
onstatState.p = onstatState.str;
lCnt = 1;
k = 0;
nl = 1;
for (; {
if (n >= l) {
l = Read(ifd, iBuf, L);
if (l < 0)
Error("read in loop 2");
iBuf[l] = 0;
if (!l)
break;
n = 0;
}
for (; n < l; ++n) {
if (tfp) {
if (nl) {
fprintf(tfp, "%d: ", lCnt);
nl = 0;
}
fputc(iBuf[n], tfp);
if (iBuf[n] == '\n')
nl = 1;
}
if (lCnt <= 3) {
if (FoundStr(&onstatState, iBuf[n])) {
if (tfp)
fprintf(tfp,
"\nstate Initialization\n");
return 1;
}
}
if (FoundStr(&onstatPrompt, iBuf[n])) {
++n;
goto loopEnd;
}
if (k > L)
Error("line too long");
oBuf[k++] = iBuf[n];
if (iBuf[n] == '\n') {
/* ignore write errors, e.g. broken pipe,
* but skip further writes to fd 1
* after an error */
if (!wErr && (e = Write(1, oBuf, k)) != k) {
wErr = 1;
if(tfp)
fprintf(tfp,
"write error, ret %d errno %d\n",
e, errno);
}
k = 0;
++lCnt;
}
}
}
loopEnd:
if (*onstatPrompt.p)
Error("eof without \"%s\" prompt", onstatPrompt.str);
if (tfp)
fprintf(tfp, "\n");

/* put any remaining iBuf contents into temp file */
if (n < l) {
if (tfp)
fprintf(tfp, "putting %d bytes into temp file\n",
l - n);
tmpFd = open(argv[2], O_WRONLY|O_CREAT|O_TRUNC, 0600);
if (tmpFd < 0)
Error("cannot open temp file %s for writing",
argv[2]);
if ((e = Write(tmpFd, iBuf + n, l - n)) != l - n)
Error("write temp file, ret %d != %d", e, l - n);
close(tmpFd);
}

if (tfp)
fprintf(tfp, "finished\n");
return 0;
}

/* Find string descriped in fstr by comparing with the next byte in the
* input stream.
*/
int
FoundStr(volatile struct findStr *fstr, int c)
{
if (!*fstr->p)
return 0;

if (*fstr->p != c)
fstr->p = fstr->str;

if (*fstr->p == c) {
++fstr->p;
return !*fstr->p;
}
else {
fstr->p = fstr->str;
return 0;
}
}

ssize_t
Read(int fd, char *buf, size_t l)
{
int e;
int rfd = fd;

if (tmpFd >= 0)
rfd = tmpFd;

do
e = read(rfd, buf, l);
while (e < 0 && errno == EINTR);

if (e != 0 || tmpFd < 0)
return e;

/* e == 0 && tmpFd >= 0, eof on temp file */
if (tfp)
fprintf(tfp, "\neof on temp file\n");
if (ftruncate(tmpFd, 0) < 0)
Error("ftruncate");
close(tmpFd);
tmpFd = -1;

do
e = read(fd, buf, l);
while (e < 0 && errno == EINTR);

return e;
}

ssize_t
Write(int fd, char *buf, size_t l)
{
int e;

do
e = write(fd, buf, l);
while (e < 0 && errno == EINTR);

return e;
}

void
handler(int sig)
{
if (tfp)
fprintf(tfp, "\nkill(pid %d, sig %d)\n", onstatPid, sig);
if (kill(onstatPid, sig))
Error("cannot kill(pid %d, sig %d)", onstatPid, sig);

/* after the signal onstat outputs a prompt without leading \n */
if (*onstatPrompt.str == '\n') {
if (onstatPrompt.p == onstatPrompt.str)
++onstatPrompt.p;
++onstatPrompt.str;
}

wErr = 1;
}

void
Error(char *str, ...)
{
va_list ap;
int e = errno;

fprintf(stderr, "cfilter: ");
va_start(ap, str);
vfprintf(stderr, str, ap);
va_end(ap);
fprintf(stderr, "\n");
if (e)
fprintf(stderr, "errno %d: %s\n", e, strerror(e));
if (tfp)
fprintf(tfp, "\nerror %d, %s\n", e, str);

exit(-1);
}

all: scansql cfilter

scansql: scansql.c
cc -O scansql.c -o scansql

cfilter: cfilter.c
cc -O cfilter.c -o cfilter

/*
* Scan sql statement text and substitute fixed texts for constats of
* variouse types:
*
* numeric: <number>
* string: <string>
* datetime: <datetime>
* interval: <interval>
*
* All text is converted into lowercase. Whitespace is reduced to one blank
* and a \n is outputtet if the line would become too long. All sql comments
* (-- and { ... }) are removed.
*
* The purpose of all this is to output the same text for two statements with
* the same structure but different constants, whitespace etc.
*
* Usage: scansql [input-file [output-file] ]
*
* input-file and output-file can be '-' for stdin and stdout
*/

#include <stdio.h>
#include <strings.h>
#include <ctype.h>
#include <stdarg.h>

/* tokens */
#define T_NONE -1
#define T_EOF 0
#define T_BRACKET_OPEN 1
#define T_BRACKET_CLOSE 2
#define T_PLUS_MINUS 3
#define T_STRING 10
#define T_NUMBER 11
#define T_IDENTIFIER 100
#define T_TO 101
#define T_DATETIME 200
#define T_INTERVAL 201
#define T_DATETIME_IV_FQUAL 202
#define T_OTHER 999

int savedToken = T_NONE;

/* these consitute one token of tpe T_OTHER */
char *op2Tab[] = {
"||",
"::",
".*",
"<=",
">=",
"<>",
"!=",
"==",
NULL
};

struct IdTab {
char *name;
int token;
};

struct IdTab idTab[] = {
"datetime", T_DATETIME,
"interval", T_INTERVAL,
"to", T_TO,
"year", T_DATETIME_IV_FQUAL,
"month", T_DATETIME_IV_FQUAL,
"day", T_DATETIME_IV_FQUAL,
"hour", T_DATETIME_IV_FQUAL,
"minute", T_DATETIME_IV_FQUAL,
"second", T_DATETIME_IV_FQUAL,
"fraction", T_DATETIME_IV_FQUAL,
NULL, T_NONE
};

FILE *ifp, *ofp;

#define MAXLINE 80 /* min output line length, \n after next token */

#define SQLTXTMAX (64*1024)
char buf[SQLTXTMAX];

char tokenBuf[SQLTXTMAX];
char putbackBuf[SQLTXTMAX];

int NextToken(char *pText);
void PutBackToken(int token, char *pText);
int ScanNumber(char **pNext, char *pText);
void ParseDatetimeInterval();
void Error(char *str, ...);

int
main(int argc, char **argv)
{
int l, token;
char *p;
int plusMinus = 0;
int lineLen = 0;

if (argc < 2 || !strcmp(argv[1], "-"))
ifp = stdin;
else {
ifp = fopen(argv[1], "r");
if (!ifp)
Error("cannot open input file %s", argv[1]);
}

if (argc < 3 || !strcmp(argv[2], "-"))
ofp = stdout;
else {
ofp = fopen(argv[2], "w");
if (!ofp)
Error("cannot open input file %s", argv[2]);
}

l = fread(buf, 1, SQLTXTMAX - 1, ifp);
if (ferror(ifp))
Error("cannot read input file %s", argv[1]);
buf[l] = 0;

/* convert to lowercase */
for (p = buf; *p; ++p)
if (isupper(*p))
*p = tolower(*p);

/* parse and output statement */
while ((token = NextToken(tokenBuf)) != T_EOF) {
/* Informix SQL: "100 + - - - - 1" is 101 !!
* Print first '+' or '-', skip following ones.
*/
if (token == T_PLUS_MINUS) {
if (plusMinus)
continue;
plusMinus = 1;
}
else
plusMinus = 0;

switch (token) {
case T_STRING:
strcpy(tokenBuf, "<string>");
break;
case T_NUMBER:
strcpy(tokenBuf, "<number>");
break;
case T_DATETIME:
ParseDatetimeInterval();
strcpy(tokenBuf, "<datetime>");
break;
case T_INTERVAL:
ParseDatetimeInterval();
strcpy(tokenBuf, "<interval>");
break;
}

l = strlen(tokenBuf);
if (lineLen > 0 && lineLen + l + 1 >= MAXLINE) {
fprintf(ofp, "\n");
lineLen = 0;
}

if (lineLen) {
fprintf(ofp, " ");
lineLen += 1;
}

fprintf(ofp, "%s", tokenBuf);
lineLen += l;
}
if (lineLen)
fprintf(ofp, "\n");

return 0;
}

void
PutBackToken(int token, char *pText)
{
savedToken = token;
strcpy(putbackBuf, pText);
}

int
NextToken(char *pText)
{
static char *next = buf;
char *p, *p2;
int c, n, l;

if (savedToken != T_NONE) {
int t = savedToken;

savedToken = T_NONE;
strcpy(pText, putbackBuf);
return t;
}

/* Skip all leading sequences of whitspace, {} amd -- comments */
do {
while (isspace(*next))
++next;

p = next;

if (next[0] == '-' && next[1] == '-') {
next += 2;
while (*next && *next++ != '\n')
;
}

if (*next == '{') {
++next;
while (*next && *next++ != '}')
;
}
} while (p != next);

if (!*next)
return T_EOF;

p = next;
if (*p == '(') {
*pText++ = *p++;
*pText = 0;
next = p;
return T_BRACKET_OPEN;
}
else if (*p == ')') {
*pText++ = *p++;
*pText = 0;
next = p;
return T_BRACKET_CLOSE;
}
else if (*p == '\'' || *p == '\"') {
c = *p;
*pText++ = *p++;
while (*p) {
*pText++ = *p++;
if (p[-1] == c)
if (*p == c)
*pText++ = *p++;
else
break;
}
*pText = 0;
next = p;
return T_STRING;
}
else if (isalpha(*p) || *p == '_') {
p2 = pText;
*pText++ = *p++;
while (isalnum(*p) || *p == '_' || *p == '$')
*pText++ = *p++;
*pText = 0;
next = p;

for (n = 0; idTab[n].name; ++n)
if (!strcmp(idTab[n].name, p2))
return idTab[n].token;

return T_IDENTIFIER;
}
else if (*p == '+' || *p == '-') {
*pText++ = *p++;
*pText = 0;
next = p;
return T_PLUS_MINUS;
}
else if (isdigit(*p)) {
return ScanNumber(&next, pText);
}
else {
for (n = 0; op2Tab[n]; ++n) {
l = strlen(op2Tab[n]);
if (!strncmp(p, op2Tab[n], l))
break;
}
if (op2Tab[n]) {
strcpy(pText, op2Tab[n]);
next = p + l;
return T_OTHER;
}

*pText++ = *p++;
*pText = 0;
next = p;
return T_OTHER;
}

Error("NextToken end");
return T_EOF; /* not reached */
}

int
ScanNumber(char **pNext, char *pText)
{
char *p = *pNext;

while (isdigit(*p))
*pText++ = *p++;

if (*p == '.') {
*pText++ = *p++;
while (isdigit(*p))
*pText++ = *p++;
}

if (*p == 'e' || *p == 'E') {
*pText++ = *p++;
if (*p == '-')
*pText++ = *p++;
else if (*p == '+')
++p;
while (isdigit(*p))
*pText++ = *p++;
}

*pText = 0;
*pNext = p;

return T_NUMBER;
}

/* datetime and interval syntax:
*
* {T_DATETIME | T_INTERVAL} T_BRACKET_OPEN ... T_BRACKET_CLOSE \
* T_DATETIME_IV_FQUAL [prec] T_TO T_DATETIME_IV_FQUAL [prec]
*
* prec: T_BRACKET_OPEN T_NUMBER T_BRACKET_CLOSE
*/
void
ParseDatetimeInterval()
{
int token;

if (NextToken(tokenBuf) != T_BRACKET_OPEN)
return;

while ((token = NextToken(tokenBuf)) != T_BRACKET_CLOSE)
if (token == T_EOF)
return;

if (NextToken(tokenBuf) != T_DATETIME_IV_FQUAL)
return;

token = NextToken(tokenBuf);
if (token == T_BRACKET_OPEN) {
if (NextToken(tokenBuf) != T_NUMBER)
return;
if (NextToken(tokenBuf) != T_BRACKET_CLOSE)
return;
token = NextToken(tokenBuf);
}

if (token != T_TO)
return;

if (NextToken(tokenBuf) != T_DATETIME_IV_FQUAL)
return;

token = NextToken(tokenBuf);
if (token == T_BRACKET_OPEN) {
if (NextToken(tokenBuf) != T_NUMBER)
return;
if (NextToken(tokenBuf) != T_BRACKET_CLOSE)
return;
}
else
PutBackToken(token, tokenBuf);
}

void
Error(char *str, ...)
{
va_list ap;

va_start(ap, str);
vfprintf(stderr, str, ap);
va_end(ap);
fprintf( stderr, "\n" );

exit(1);
}

i get a directory checksum error.
i downloaded your file to my laptop (windows xp)
and then ftp that to my dev UNIX HP 11i server.


Tom

Tom,

I assume you made some mistake in transferring the file (like using ftp
text mode?). I tried on HP myself without problems:

$ uname -a
HP-UX trift B.11.11 U 9000/800 832960671 unlimited-user license

$ ls -l
total 32
-rw-r--r-- 1 michaelm ts 15270 Apr 19 17:03 sqlstat.tar.gz

$ sum sqlstat.tar.gz
48564 30 sqlstat.tar.gz

$ gunzip sqlstat.tar.gz
$ tar tvf sqlstat.tar
rwxr-xr-x 1000/100 0 Mar 5 21:47 2005 sqlstat/
rw-r--r-- 1000/100 18007 Jul 21 22:43 2004 sqlstat/COPYING
rwxr-xr-x 1000/100 7571 Mar 5 21:47 2005 sqlstat/SqlStat
rwxr-xr-x 1000/100 0 Mar 4 22:17 2005 sqlstat/erg.21:25:37/
rw-r--r-- 1000/100 93 Mar 4 22:17 2005 sqlstat/erg.21:25:37/INFO
rw-r--r-- 1000/100 7726 Jul 21 22:41 2004
sqlstat/erg.21:25:37/00083.16476_1
rw-r--r-- 1000/100 1750 Jul 21 22:41 2004
sqlstat/erg.21:25:37/00018.16460_1
rw-r--r-- 1000/100 1231 Jul 21 22:41 2004
sqlstat/erg.21:25:37/00013.35489_1
rw-r--r-- 1000/100 7386 Mar 5 21:47 2005 sqlstat/cfilter.c
rw-r--r-- 1000/100 6961 Mar 4 22:13 2005 sqlstat/scansql.c
rw-r--r-- 1000/100 120 Mar 4 22:13 2005 sqlstat/makefile

Can you verify the file size (15270) and checksum (48564 30)?

Michael


tomcaml@yahoo.com wrote:
> i get a directory checksum error.
> i downloaded your file to my laptop (windows xp)
> and then ftp that to my dev UNIX HP 11i server.
>
>
> Tom
>

Michael,

When i download it from your web site, the file says sqlstat.tar.tar,
not sqlstat.tar.gz as you have it.

$ sum sqlstat.tar.tar
48564 30 sqlstat.tar.tar

$ ls -l
-rw-r----- 1 informix informix 15270 Apr 19 09:27
sqlstat.tar.tar

$ uname -a
HP-UX B.11.11 U 9000/800 1488063140 unlimited-user license


Thanks,
Tom

tomcaml@yahoo.com wrote:
> Michael,
>
> When i download it from your web site, the file says sqlstat.tar.tar,
> not sqlstat.tar.gz as you have it.
>
> $ sum sqlstat.tar.tar
> 48564 30 sqlstat.tar.tar
>
> $ ls -l
> -rw-r----- 1 informix informix 15270 Apr 19 09:27
> sqlstat.tar.tar
>
> $ uname -a
> HP-UX B.11.11 U 9000/800 1488063140 unlimited-user license
>
>
> Thanks,
> Tom
>


Tom,

this looks fine so far, except for the name of the file (presumably
messed up by your download tool). I assume you then tried

tar xf sqlstat.tar.tar

and got an error. Try this:

file sqlstat.tar.tar # output: gzip compressed, not tar file!
mv sqlstat.tar.tar sqlstat.tar.gz
gunzip sqlstat.tar.gz
tar xf sqlstat.tar

Michael