Re: Bug in expr***urgent

[salih k]

On Tue, Jan 19, 2010 at 9:10 PM, Eric Blake <address@hidden> wrote:

> According to salih k on 1/16/2010 7:32 AM:
> > Piece of Script
> > --------------------------
> >
> >      /isnum=`awk -F$delim '$1=="BH"{print $5}' $fil`/
> >     /
>
> That's still not exactly what you ran (you marked it up afterwards), but
> it is close enough, I suppose.
>
> >             int_num=`echo -e $isnum | cut -f1 -d'.'`
>
> echo -e is not portable.  And if all you are trying to do is grab
> everything before the first '.', you can do that without even forking:
>
> int_num=${isnum%%.*}
>
> But using a more efficient implementation won't change the net result of
> your script, at least on systems where echo -e does what you wanted.
>
> >             add_num=`expr $int_num + 1 1>/dev/null 2>&1`
>
> This line sets add_num to the empty string (because there is no output
> from expr).  Why run a command substitution if there won't be any output?
> Also, your lack of shell quoting could cause you grief; it would be more
> robust to use "$int_num" rather than $int_num.  The following is similar
> to your above line, but safer:
>
> add_num=
> expr "$int_num" + 1 >/dev/null 2>&1
> kms=$?
>
> And even more efficiently, you can get the same results without forking,
> by following my earlier advice:
>
> case $int_num in
>  *[!0-9]*) kms=1; echo non-numeric;;
>  *) kms=0; echo numeric;;
> esac
>
> But again, switching to a more efficient implementation shouldn't seem to
> have any bearing on your results.
>
> >      kms=$?
> >            # if [ "$?" -ne "0" ]
> >              if [ "$kms" -ne "0" ]
> >             then
> >             log_msg "Debug msg4.new1 File has expr pstat $kms int_num is
> >     $int_num"
>
> Well, I don't see anything obvious about why you are seeing this:
>
> > Debug msg3.1 File FILE1 has int_num1 28 has ps 0
> > Debug msg4.new1 File has expr pstat 51198 int_num is 28
> ...
> > Debug msg3.1 File  FILE2 has int_num1 1  has ps 0
>
> then again, you post-formatted the output, too (notice the inconsistent
> spacing), and you didn't post the definition of log_msg.  So who knows
> what else your post-formatting has changed, such that we aren't seeing
> EXACTLY what happened to you.
>
> But if it is a bug, then the bug is in your /bin/sh and not in expr.  And
> if you really do have a buggy shell, then there's nothing this list can do
> to help you.
>
> > Is it because add_num=`expr $int_num + *1 1>/dev*/null 2>&1` whether i
> > have to chanage it to
> > add_num=`expr $int_num + *1 >*/dev/null 2>&1`?
>
> The change between '>/dev/null' and '1>/dev/null' has no effect - they are
> strictly equivalent in shell programming.  You may be benefited by doing
> more homework on shell programming.  But so far, you have failed to show
> any evidence of a bug in expr.  And inserting spurious * into the lines
> doesn't make your example any easier to attempt and reproduce.
>
> --
>  Don't work too hard, make some time for fun as well!
>
> Eric Blake             address@hidden
>
>
Hi,
Thats fine but I did not got the answer why the exit status shows more than
255 during first run ,
ie exit status is 51198 even though the argument to expr int_num is 28
and in some cases it is 1552 even though the argument to expr int_num is 1.
To show this one i have provided the log

*> Debug msg3.1 File FILE1 has int_num1 28 has ps 0
> Debug msg4.new1 File has expr pstat 51198 int_num is 28
...
*
*> Debug msg3.1 File  FILE2 has int_num1 1  has ps 0*
*
function *log_msg
log_msg()
{
        echo -e $* | tee -a $LOG
}
LOG=$LOG_DIR
and $BASE/log--afile name from thios i provided the log
So you are saying that this will not happen but this happened that y the
log.What else I can provide.
As a solution am asking whether I can check *if [ $kms -eq 1 -o $kms –eq 2
–o $kms –eq 3 ] *instaed of *if [ "$kms" -ne "0" ]*
 as the mannual says
*info expr|more
    Exit status:*
*      0 if the expression is neither null nor 0,
      1 if the expression is null or 0,
      2 if the expression is syntactically invalid,
      3 if an error occurred.*
as the manager is not happy to change the command (  case $int_num in
*[!0-9]*|'') non nume ;;))as this was working in unix and also this issue
happens in rare.we don t think issue in script.The rest of the code is for
different purpose if u want that also i dont know what advantage we will
get.The same files loaded without any issue in the second time.!!!
Please advise

Thanks,
Salih